.. currentmodule:: sklearn

===============
Release history
===============

Version 0.19.1
==============

**October, 2017**

This is a bug-fix release with some minor documentation improvements and
enhancements to features released in 0.19.0.

Note there may be minor differences in TSNE output in this release (due to
:issue:`9623`), in the case where multiple samples have equal distance to some
sample.

Changelog
---------

API changes
...........

- Reverted the addition of ``metrics.ndcg_score`` and ``metrics.dcg_score``
  which had been merged into version 0.19.0 by error.  The implementations
  were broken and undocumented.

- ``return_train_score`` which was added to
  :class:`model_selection.GridSearchCV`,
  :class:`model_selection.RandomizedSearchCV` and
  :func:`model_selection.cross_validate` in version 0.19.0 will be changing its
  default value from True to False in version 0.21.  We found that calculating
  training score could have a great effect on cross validation runtime in some
  cases.  Users should explicitly set ``return_train_score`` to False if
  prediction or scoring functions are slow, resulting in a deleterious effect
  on CV runtime, or to True if they wish to use the calculated scores.
  :issue:`9677` by :user:`Kumar Ashutosh <thechargedneutron>` and `Joel
  Nothman`_.

- ``correlation_models`` and ``regression_models`` from the legacy gaussian
  processes implementation have been belatedly deprecated. :issue:`9717` by
  :user:`Kumar Ashutosh <thechargedneutron>`.

Bug fixes
.........

- Avoid integer overflows in :func:`metrics.matthews_corrcoef`.
  :issue:`9693` by :user:`Sam Steingold <sam-s>`.

- Fix ValueError in :class:`preprocessing.LabelEncoder` when using
  ``inverse_transform`` on unseen labels. :issue:`9816` by :user:`Charlie Newey
  <newey01c>`.

- Fixed a bug in the objective function for :class:`manifold.TSNE` (both exact
  and with the Barnes-Hut approximation) when ``n_components >= 3``.
  :issue:`9711` by :user:`goncalo-rodrigues`.

- Fix regression in :func:`model_selection.cross_val_predict` where it
  raised an error with ``method='predict_proba'`` for some probabilistic
  classifiers. :issue:`9641` by :user:`James Bourbeau <jrbourbeau>`.

- Fixed a bug where :func:`datasets.make_classification` modified its input
  ``weights``. :issue:`9865` by :user:`Sachin Kelkar <s4chin>`.

- :class:`model_selection.StratifiedShuffleSplit` now works with multioutput
  multiclass or multilabel data with more than 1000 columns.  :issue:`9922` by
  :user:`Charlie Brummitt <crbrummitt>`.

- Fixed a bug with nested and conditional parameter setting, e.g. setting a
  pipeline step and its parameter at the same time. :issue:`9945` by `Andreas
  Müller`_ and `Joel Nothman`_.

Regressions in 0.19.0 fixed in 0.19.1:

- Fixed a bug where parallelised prediction in random forests was not
  thread-safe and could (rarely) result in arbitrary errors. :issue:`9830` by
  `Joel Nothman`_.

- Fix regression in :func:`model_selection.cross_val_predict` where it no
  longer accepted ``X`` as a list. :issue:`9600` by :user:`Rasul Kerimov
  <CoderINusE>`.

- Fixed handling of :func:`cross_val_predict` for binary classification with
  ``method='decision_function'``. :issue:`9593` by :user:`Reiichiro Nakano
  <reiinakano>` and core devs.

- Fix regression in :class:`pipeline.Pipeline` where it no longer accepted
  ``steps`` as a tuple. :issue:`9604` by :user:`Joris Van den Bossche
  <jorisvandenbossche>`.

- Fix bug where ``n_iter`` was not properly deprecated, leaving ``n_iter``
  unavailable for interim use in
  :class:`linear_model.SGDClassifier`, :class:`linear_model.SGDRegressor`,
  :class:`linear_model.PassiveAggressiveClassifier`,
  :class:`linear_model.PassiveAggressiveRegressor` and
  :class:`linear_model.Perceptron`. :issue:`9558` by `Andreas Müller`_.

- Dataset fetchers make sure temporary files are closed before removing them,
  which caused errors on Windows. :issue:`9847` by :user:`Joan Massich <massich>`.

- Fixed a regression in :class:`manifold.TSNE` where it no longer supported
  metrics other than 'euclidean' and 'precomputed'. :issue:`9623` by :user:`Oli
  Blum <oliblum90>`.

Enhancements
............

- Our test suite and :func:`utils.estimator_checks.check_estimators` can now be
  run without Nose installed. :issue:`9697` by :user:`Joan Massich <massich>`.

- To improve usability of version 0.19's :class:`pipeline.Pipeline`
  caching, ``memory`` now allows ``joblib.Memory`` instances.
  This make use of the new :func:`utils.validation.check_memory` helper.
  issue:`9584` by :user:`Kumar Ashutosh <thechargedneutron>`

- Some fixes to examples: :issue:`9750`, :issue:`9788`, :issue:`9815`

- Made a FutureWarning in SGD-based estimators less verbose. :issue:`9802` by
  :user:`Vrishank Bhardwaj <vrishank97>`.

Code and Documentation Contributors
-----------------------------------

With thanks to:

Joel Nothman, Loic Esteve, Andreas Mueller, Kumar Ashutosh,
Vrishank Bhardwaj, Hanmin Qin, Rasul Kerimov, James Bourbeau,
Nagarjuna Kumar, Nathaniel Saul, Olivier Grisel, Roman
Yurchak, Reiichiro Nakano, Sachin Kelkar, Sam Steingold,
Yaroslav Halchenko, diegodlh, felix, goncalo-rodrigues,
jkleint, oliblum90, pasbi, Anthony Gitter, Ben Lawson, Charlie
Brummitt, Didi Bar-Zev, Gael Varoquaux, Joan Massich, Joris
Van den Bossche, nielsenmarkus11

Version 0.19
============

**August 12, 2017**

Highlights
----------

We are excited to release a number of great new features including
:class:`neighbors.LocalOutlierFactor` for anomaly detection,
:class:`preprocessing.QuantileTransformer` for robust feature transformation,
and the :class:`multioutput.ClassifierChain` meta-estimator to simply account
for dependencies between classes in multilabel problems. We have some new
algorithms in existing estimators, such as multiplicative update in
:class:`decomposition.NMF` and multinomial
:class:`linear_model.LogisticRegression` with L1 loss (use ``solver='saga'``).

Cross validation is now able to return the results from multiple metric
evaluations. The new :func:`model_selection.cross_validate` can return many
scores on the test data as well as training set performance and timings, and we
have extended the ``scoring`` and ``refit`` parameters for grid/randomized
search :ref:`to handle multiple metrics <multimetric_grid_search>`.

You can also learn faster.  For instance, the :ref:`new option to cache
transformations <pipeline_cache>` in :class:`pipeline.Pipeline` makes grid
search over pipelines including slow transformations much more efficient.  And
you can predict faster: if you're sure you know what you're doing, you can turn
off validating that the input is finite using :func:`config_context`.

We've made some important fixes too.  We've fixed a longstanding implementation
error in :func:`metrics.average_precision_score`, so please be cautious with
prior results reported from that function.  A number of errors in the
:class:`manifold.TSNE` implementation have been fixed, particularly in the
default Barnes-Hut approximation.  :class:`semi_supervised.LabelSpreading` and
:class:`semi_supervised.LabelPropagation` have had substantial fixes.
LabelPropagation was previously broken. LabelSpreading should now correctly
respect its alpha parameter.

Changed models
--------------

The following estimators and functions, when fit with the same data and
parameters, may produce different models from the previous version. This often
occurs due to changes in the modelling logic (bug fixes or enhancements), or in
random sampling procedures.

- :class:`cluster.KMeans` with sparse X and initial centroids given (bug fix)
- :class:`cross_decomposition.PLSRegression`
  with ``scale=True`` (bug fix)
- :class:`ensemble.GradientBoostingClassifier` and
  :class:`ensemble.GradientBoostingRegressor` where ``min_impurity_split`` is used (bug fix)
- gradient boosting ``loss='quantile'`` (bug fix)
- :class:`ensemble.IsolationForest` (bug fix)
- :class:`feature_selection.SelectFdr` (bug fix)
- :class:`linear_model.RANSACRegressor` (bug fix)
- :class:`linear_model.LassoLars` (bug fix)
- :class:`linear_model.LassoLarsIC` (bug fix)
- :class:`manifold.TSNE` (bug fix)
- :class:`neighbors.NearestCentroid` (bug fix)
- :class:`semi_supervised.LabelSpreading` (bug fix)
- :class:`semi_supervised.LabelPropagation` (bug fix)
- tree based models where ``min_weight_fraction_leaf`` is used (enhancement)

Details are listed in the changelog below.

(While we are trying to better inform users by providing this information, we
cannot assure that this list is complete.)

Changelog
---------

New features
............

Classifiers and regressors

- Added :class:`multioutput.ClassifierChain` for multi-label
  classification. By `Adam Kleczewski <adamklec>`_.

- Added solver ``'saga'`` that implements the improved version of Stochastic
  Average Gradient, in :class:`linear_model.LogisticRegression` and
  :class:`linear_model.Ridge`. It allows the use of L1 penalty with
  multinomial logistic loss, and behaves marginally better than 'sag'
  during the first epochs of ridge and logistic regression.
  :issue:`8446` by `Arthur Mensch`_.

Other estimators

- Added the :class:`neighbors.LocalOutlierFactor` class for anomaly
  detection based on nearest neighbors.
  :issue:`5279` by `Nicolas Goix`_ and `Alexandre Gramfort`_.

- Added :class:`preprocessing.QuantileTransformer` class and
  :func:`preprocessing.quantile_transform` function for features
  normalization based on quantiles.
  :issue:`8363` by :user:`Denis Engemann <dengemann>`,
  :user:`Guillaume Lemaitre <glemaitre>`, `Olivier Grisel`_, `Raghav RV`_,
  :user:`Thierry Guillemot <tguillemot>`, and `Gael Varoquaux`_.

- The new solver ``'mu'`` implements a Multiplicate Update in
  :class:`decomposition.NMF`, allowing the optimization of all
  beta-divergences, including the Frobenius norm, the generalized
  Kullback-Leibler divergence and the Itakura-Saito divergence.
  :issue:`5295` by `Tom Dupre la Tour`_.

Model selection and evaluation

- :class:`model_selection.GridSearchCV` and
  :class:`model_selection.RandomizedSearchCV` now support simultaneous
  evaluation of multiple metrics. Refer to the
  :ref:`multimetric_grid_search` section of the user guide for more
  information. :issue:`7388` by `Raghav RV`_

- Added the :func:`model_selection.cross_validate` which allows evaluation
  of multiple metrics. This function returns a dict with more useful
  information from cross-validation such as the train scores, fit times and
  score times.
  Refer to :ref:`multimetric_cross_validation` section of the userguide
  for more information. :issue:`7388` by `Raghav RV`_

- Added :func:`metrics.mean_squared_log_error`, which computes
  the mean square error of the logarithmic transformation of targets,
  particularly useful for targets with an exponential trend.
  :issue:`7655` by :user:`Karan Desai <karandesai-96>`.

- Added :func:`metrics.dcg_score` and :func:`metrics.ndcg_score`, which
  compute Discounted cumulative gain (DCG) and Normalized discounted
  cumulative gain (NDCG).
  :issue:`7739` by :user:`David Gasquez <davidgasquez>`.

- Added the :class:`model_selection.RepeatedKFold` and
  :class:`model_selection.RepeatedStratifiedKFold`.
  :issue:`8120` by `Neeraj Gangwar`_.

- Added a scorer based on :class:`metrics.explained_variance_score`.
  :issue:`9259` by `Hanmin Qin <https://github.com/qinhanmin2014>`_. 

Miscellaneous

- Validation that input data contains no NaN or inf can now be suppressed
  using :func:`config_context`, at your own risk. This will save on runtime,
  and may be particularly useful for prediction time. :issue:`7548` by
  `Joel Nothman`_.

- Added a test to ensure parameter listing in docstrings match the
  function/class signature. :issue:`9206` by `Alexandre Gramfort`_ and
  `Raghav RV`_.

Enhancements
............

Trees and ensembles

- The ``min_weight_fraction_leaf`` constraint in tree construction is now
  more efficient, taking a fast path to declare a node a leaf if its weight
  is less than 2 * the minimum. Note that the constructed tree will be
  different from previous versions where ``min_weight_fraction_leaf`` is
  used. :issue:`7441` by :user:`Nelson Liu <nelson-liu>`.

- :class:`ensemble.GradientBoostingClassifier` and :class:`ensemble.GradientBoostingRegressor`
  now support sparse input for prediction.
  :issue:`6101` by :user:`Ibraim Ganiev <olologin>`.

- :class:`ensemble.VotingClassifier` now allows changing estimators by using
  :meth:`ensemble.VotingClassifier.set_params`. An estimator can also be
  removed by setting it to ``None``.
  :issue:`7674` by :user:`Yichuan Liu <yl565>`.

- :func:`tree.export_graphviz` now shows configurable number of decimal
  places. :issue:`8698` by :user:`Guillaume Lemaitre <glemaitre>`.

- Added ``flatten_transform`` parameter to :class:`ensemble.VotingClassifier`
  to change output shape of `transform` method to 2 dimensional.
  :issue:`7794` by :user:`Ibraim Ganiev <olologin>` and
  :user:`Herilalaina Rakotoarison <herilalaina>`.

Linear, kernelized and related models

- :class:`linear_model.SGDClassifier`, :class:`linear_model.SGDRegressor`,
  :class:`linear_model.PassiveAggressiveClassifier`,
  :class:`linear_model.PassiveAggressiveRegressor` and
  :class:`linear_model.Perceptron` now expose ``max_iter`` and
  ``tol`` parameters, to handle convergence more precisely.
  ``n_iter`` parameter is deprecated, and the fitted estimator exposes
  a ``n_iter_`` attribute, with actual number of iterations before
  convergence. :issue:`5036` by `Tom Dupre la Tour`_.

- Added ``average`` parameter to perform weight averaging in
  :class:`linear_model.PassiveAggressiveClassifier`. :issue:`4939`
  by :user:`Andrea Esuli <aesuli>`.

- :class:`linear_model.RANSACRegressor` no longer throws an error
  when calling ``fit`` if no inliers are found in its first iteration.
  Furthermore, causes of skipped iterations are tracked in newly added
  attributes, ``n_skips_*``.
  :issue:`7914` by :user:`Michael Horrell <mthorrell>`.

- In :class:`gaussian_process.GaussianProcessRegressor`, method ``predict``
  is a lot faster with ``return_std=True``. :issue:`8591` by
  :user:`Hadrien Bertrand <hbertrand>`.

- Added ``return_std`` to ``predict`` method of
  :class:`linear_model.ARDRegression` and
  :class:`linear_model.BayesianRidge`.
  :issue:`7838` by :user:`Sergey Feldman <sergeyf>`.

- Memory usage enhancements: Prevent cast from float32 to float64 in:
  :class:`linear_model.MultiTaskElasticNet`;
  :class:`linear_model.LogisticRegression` when using newton-cg solver; and
  :class:`linear_model.Ridge` when using svd, sparse_cg, cholesky or lsqr
  solvers. :issue:`8835`, :issue:`8061` by :user:`Joan Massich <massich>` and :user:`Nicolas
  Cordier <ncordier>` and :user:`Thierry Guillemot <tguillemot>`.

Other predictors

- Custom metrics for the :mod:`neighbors` binary trees now have
  fewer constraints: they must take two 1d-arrays and return a float.
  :issue:`6288` by `Jake Vanderplas`_.

- ``algorithm='auto`` in :mod:`neighbors` estimators now chooses the most
  appropriate algorithm for all input types and metrics. :issue:`9145` by
  :user:`Herilalaina Rakotoarison <herilalaina>` and :user:`Reddy Chinthala
  <preddy5Pradyumna>`.

Decomposition, manifold learning and clustering

- :class:`cluster.MiniBatchKMeans` and :class:`cluster.KMeans`
  now use significantly less memory when assigning data points to their
  nearest cluster center. :issue:`7721` by :user:`Jon Crall <Erotemic>`.

- :class:`decomposition.PCA`, :class:`decomposition.IncrementalPCA` and
  :class:`decomposition.TruncatedSVD` now expose the singular values
  from the underlying SVD. They are stored in the attribute
  ``singular_values_``, like in :class:`decomposition.IncrementalPCA`.
  :issue:`7685` by :user:`Tommy Löfstedt <tomlof>`

- Fixed the implementation of noise_variance_ in :class:`decomposition.PCA`.
  :issue:`9108` by `Hanmin Qin <https://github.com/qinhanmin2014>`_.

- :class:`decomposition.NMF` now faster when ``beta_loss=0``.
  :issue:`9277` by :user:`hongkahjun`.

- Memory improvements for method ``barnes_hut`` in :class:`manifold.TSNE`
  :issue:`7089` by :user:`Thomas Moreau <tomMoral>` and `Olivier Grisel`_.

- Optimization schedule improvements for Barnes-Hut :class:`manifold.TSNE`
  so the results are closer to the one from the reference implementation
  `lvdmaaten/bhtsne <https://github.com/lvdmaaten/bhtsne>`_ by :user:`Thomas
  Moreau <tomMoral>` and `Olivier Grisel`_.

- Memory usage enhancements: Prevent cast from float32 to float64 in
  :class:`decomposition.PCA` and
  :func:`decomposition.randomized_svd_low_rank`.
  :issue:`9067` by `Raghav RV`_.

Preprocessing and feature selection

- Added ``norm_order`` parameter to :class:`feature_selection.SelectFromModel`
  to enable selection of the norm order when ``coef_`` is more than 1D.
  :issue:`6181` by :user:`Antoine Wendlinger <antoinewdg>`.

- Added ability to use sparse matrices in :func:`feature_selection.f_regression`
  with ``center=True``. :issue:`8065` by :user:`Daniel LeJeune <acadiansith>`.

- Small performance improvement to n-gram creation in
  :mod:`feature_extraction.text` by binding methods for loops and
  special-casing unigrams. :issue:`7567` by :user:`Jaye Doepke <jtdoepke>`

- Relax assumption on the data for the
  :class:`kernel_approximation.SkewedChi2Sampler`. Since the Skewed-Chi2
  kernel is defined on the open interval :math:`(-skewedness; +\infty)^d`,
  the transform function should not check whether ``X < 0`` but whether ``X <
  -self.skewedness``. :issue:`7573` by :user:`Romain Brault <RomainBrault>`.

- Made default kernel parameters kernel-dependent in
  :class:`kernel_approximation.Nystroem`.
  :issue:`5229` by :user:`Saurabh Bansod <mth4saurabh>` and `Andreas Müller`_.

Model evaluation and meta-estimators

- :class:`pipeline.Pipeline` is now able to cache transformers
  within a pipeline by using the ``memory`` constructor parameter.
  :issue:`7990` by :user:`Guillaume Lemaitre <glemaitre>`.

- :class:`pipeline.Pipeline` steps can now be accessed as attributes of its
  ``named_steps`` attribute. :issue:`8586` by :user:`Herilalaina
  Rakotoarison <herilalaina>`.

- Added ``sample_weight`` parameter to :meth:`pipeline.Pipeline.score`.
  :issue:`7723` by :user:`Mikhail Korobov <kmike>`.

- Added ability to set ``n_jobs`` parameter to :func:`pipeline.make_union`.
  A ``TypeError`` will be raised for any other kwargs. :issue:`8028`
  by :user:`Alexander Booth <alexandercbooth>`.

- :class:`model_selection.GridSearchCV`,
  :class:`model_selection.RandomizedSearchCV` and
  :func:`model_selection.cross_val_score` now allow estimators with callable
  kernels which were previously prohibited.
  :issue:`8005` by `Andreas Müller`_ .

- :func:`model_selection.cross_val_predict` now returns output of the
  correct shape for all values of the argument ``method``.
  :issue:`7863` by :user:`Aman Dalmia <dalmia>`.

- Added ``shuffle`` and ``random_state`` parameters to shuffle training
  data before taking prefixes of it based on training sizes in
  :func:`model_selection.learning_curve`.
  :issue:`7506` by :user:`Narine Kokhlikyan <NarineK>`.

- :class:`model_selection.StratifiedShuffleSplit` now works with multioutput
  multiclass (or multilabel) data.  :issue:`9044` by `Vlad Niculae`_.

- Speed improvements to :class:`model_selection.StratifiedShuffleSplit`.
  :issue:`5991` by :user:`Arthur Mensch <arthurmensch>` and `Joel Nothman`_.

- Add ``shuffle`` parameter to :func:`model_selection.train_test_split`.
  :issue:`8845` by  :user:`themrmax <themrmax>`

- :class:`multioutput.MultiOutputRegressor` and :class:`multioutput.MultiOutputClassifier`
  now support online learning using ``partial_fit``.
  :issue: `8053` by :user:`Peng Yu <yupbank>`.

- Add ``max_train_size`` parameter to :class:`model_selection.TimeSeriesSplit`
  :issue:`8282` by :user:`Aman Dalmia <dalmia>`.

- More clustering metrics are now available through :func:`metrics.get_scorer`
  and ``scoring`` parameters. :issue:`8117` by `Raghav RV`_.

Metrics

- :func:`metrics.matthews_corrcoef` now support multiclass classification.
  :issue:`8094` by :user:`Jon Crall <Erotemic>`.

- Add ``sample_weight`` parameter to :func:`metrics.cohen_kappa_score`.
  :issue:`8335` by :user:`Victor Poughon <vpoughon>`.

Miscellaneous

- :func:`utils.check_estimator` now attempts to ensure that methods
  transform, predict, etc.  do not set attributes on the estimator.
  :issue:`7533` by :user:`Ekaterina Krivich <kiote>`.

- Added type checking to the ``accept_sparse`` parameter in
  :mod:`utils.validation` methods. This parameter now accepts only boolean,
  string, or list/tuple of strings. ``accept_sparse=None`` is deprecated and
  should be replaced by ``accept_sparse=False``.
  :issue:`7880` by :user:`Josh Karnofsky <jkarno>`.

- Make it possible to load a chunk of an svmlight formatted file by
  passing a range of bytes to :func:`datasets.load_svmlight_file`.
  :issue:`935` by :user:`Olivier Grisel <ogrisel>`.

- :class:`dummy.DummyClassifier` and :class:`dummy.DummyRegressor`
  now accept non-finite features. :issue:`8931` by :user:`Attractadore`.

Bug fixes
.........

Trees and ensembles

- Fixed a memory leak in trees when using trees with ``criterion='mae'``.
  :issue:`8002` by `Raghav RV`_.

- Fixed a bug where :class:`ensemble.IsolationForest` uses an
  an incorrect formula for the average path length
  :issue:`8549` by `Peter Wang <https://github.com/PTRWang>`_.

- Fixed a bug where :class:`ensemble.AdaBoostClassifier` throws
  ``ZeroDivisionError`` while fitting data with single class labels.
  :issue:`7501` by :user:`Dominik Krzeminski <dokato>`.

- Fixed a bug in :class:`ensemble.GradientBoostingClassifier` and
  :class:`ensemble.GradientBoostingRegressor` where a float being compared
  to ``0.0`` using ``==`` caused a divide by zero error. :issue:`7970` by
  :user:`He Chen <chenhe95>`.

- Fix a bug where :class:`ensemble.GradientBoostingClassifier` and
  :class:`ensemble.GradientBoostingRegressor` ignored the
  ``min_impurity_split`` parameter.
  :issue:`8006` by :user:`Sebastian Pölsterl <sebp>`.

- Fixed ``oob_score`` in :class:`ensemble.BaggingClassifier`.
  :issue:`8936` by :user:`Michael Lewis <mlewis1729>`

- Fixed excessive memory usage in prediction for random forests estimators.
  :issue:`8672` by :user:`Mike Benfield <mikebenfield>`.

- Fixed a bug where ``sample_weight`` as a list broke random forests in Python 2
  :issue:`8068` by :user:`xor`.

- Fixed a bug where :class:`ensemble.IsolationForest` fails when
  ``max_features`` is less than 1.
  :issue:`5732` by :user:`Ishank Gulati <IshankGulati>`.

- Fix a bug where gradient boosting with ``loss='quantile'`` computed
  negative errors for negative values of ``ytrue - ypred`` leading to wrong
  values when calling ``__call__``.
  :issue:`8087` by :user:`Alexis Mignon <AlexisMignon>`

- Fix a bug where :class:`ensemble.VotingClassifier` raises an error
  when a numpy array is passed in for weights. :issue:`7983` by
  :user:`Vincent Pham <vincentpham1991>`.

- Fixed a bug where :func:`tree.export_graphviz` raised an error
  when the length of features_names does not match n_features in the decision
  tree. :issue:`8512` by :user:`Li Li <aikinogard>`.

Linear, kernelized and related models

- Fixed a bug where :func:`linear_model.RANSACRegressor.fit` may run until
  ``max_iter`` if it finds a large inlier group early. :issue:`8251` by
  :user:`aivision2020`.

- Fixed a bug where :class:`naive_bayes.MultinomialNB` and
  :class:`naive_bayes.BernoulliNB` failed when ``alpha=0``. :issue:`5814` by
  :user:`Yichuan Liu <yl565>` and :user:`Herilalaina Rakotoarison
  <herilalaina>`.

- Fixed a bug where :class:`linear_model.LassoLars` does not give
  the same result as the LassoLars implementation available
  in R (lars library). :issue:`7849` by :user:`Jair Montoya Martinez <jmontoyam>`.

- Fixed a bug in :class:`linear_model.RandomizedLasso`,
  :class:`linear_model.Lars`, :class:`linear_model.LassoLars`,
  :class:`linear_model.LarsCV` and :class:`linear_model.LassoLarsCV`,
  where the parameter ``precompute`` was not used consistently across
  classes, and some values proposed in the docstring could raise errors.
  :issue:`5359` by `Tom Dupre la Tour`_.

- Fix inconsistent results between :class:`linear_model.RidgeCV` and
  :class:`linear_model.Ridge` when using ``normalize=True``. :issue:`9302`
  by `Alexandre Gramfort`_.

- Fix a bug where :func:`linear_model.LassoLars.fit` sometimes
  left ``coef_`` as a list, rather than an ndarray.
  :issue:`8160` by :user:`CJ Carey <perimosocordiae>`.

- Fix :func:`linear_model.BayesianRidge.fit` to return
  ridge parameter ``alpha_`` and ``lambda_`` consistent with calculated
  coefficients ``coef_`` and ``intercept_``.
  :issue:`8224` by :user:`Peter Gedeck <gedeck>`.

- Fixed a bug in :class:`svm.OneClassSVM` where it returned floats instead of
  integer classes. :issue:`8676` by :user:`Vathsala Achar <VathsalaAchar>`.

- Fix AIC/BIC criterion computation in :class:`linear_model.LassoLarsIC`.
  :issue:`9022` by `Alexandre Gramfort`_ and :user:`Mehmet Basbug <mehmetbasbug>`.

- Fixed a memory leak in our LibLinear implementation. :issue:`9024` by
  :user:`Sergei Lebedev <superbobry>`

- Fix bug where stratified CV splitters did not work with
  :class:`linear_model.LassoCV`. :issue:`8973` by
  :user:`Paulo Haddad <paulochf>`.

- Fixed a bug in :class:`gaussian_process.GaussianProcessRegressor`
  when the standard deviation and covariance predicted without fit
  would fail with a unmeaningful error by default.
  :issue:`6573` by :user:`Quazi Marufur Rahman <qmaruf>` and
  `Manoj Kumar`_.

Other predictors

- Fix :class:`semi_supervised.BaseLabelPropagation` to correctly implement
  ``LabelPropagation`` and ``LabelSpreading`` as done in the referenced
  papers. :issue:`9239`
  by :user:`Andre Ambrosio Boechat <boechat107>`, :user:`Utkarsh Upadhyay
  <musically-ut>`, and `Joel Nothman`_.

Decomposition, manifold learning and clustering

- Fixed the implementation of :class:`manifold.TSNE`:
- ``early_exageration`` parameter had no effect and is now used for the
  first 250 optimization iterations.
- Fixed the ``AssertionError: Tree consistency failed`` exception
  reported in :issue:`8992`.
- Improve the learning schedule to match the one from the reference
  implementation `lvdmaaten/bhtsne <https://github.com/lvdmaaten/bhtsne>`_.
     by :user:`Thomas Moreau <tomMoral>` and `Olivier Grisel`_.

- Fix a bug in :class:`decomposition.LatentDirichletAllocation`
  where the ``perplexity`` method was returning incorrect results because
  the ``transform`` method returns normalized document topic distributions
  as of version 0.18. :issue:`7954` by :user:`Gary Foreman <garyForeman>`.

- Fix output shape and bugs with n_jobs > 1 in
  :class:`decomposition.SparseCoder` transform and
  :func:`decomposition.sparse_encode`
  for one-dimensional data and one component.
  This also impacts the output shape of :class:`decomposition.DictionaryLearning`.
  :issue:`8086` by `Andreas Müller`_.

- Fixed the implementation of ``explained_variance_``
  in :class:`decomposition.PCA`,
  :class:`decomposition.RandomizedPCA` and
  :class:`decomposition.IncrementalPCA`.
  :issue:`9105` by `Hanmin Qin <https://github.com/qinhanmin2014>`_.

- Fixed the implementation of noise_variance_ in :class:`decomposition.PCA`.
  :issue:`9108` by `Hanmin Qin <https://github.com/qinhanmin2014>`_.

- Fixed a bug where :class:`cluster.DBSCAN` gives incorrect
  result when input is a precomputed sparse matrix with initial
  rows all zero. :issue:`8306` by :user:`Akshay Gupta <Akshay0724>`

- Fix a bug regarding fitting :class:`cluster.KMeans` with a sparse
  array X and initial centroids, where X's means were unnecessarily being
  subtracted from the centroids. :issue:`7872` by :user:`Josh Karnofsky <jkarno>`.

- Fixes to the input validation in :class:`covariance.EllipticEnvelope`.
  :issue:`8086` by `Andreas Müller`_.

- Fixed a bug in :class:`covariance.MinCovDet` where inputting data
  that produced a singular covariance matrix would cause the helper method
  ``_c_step`` to throw an exception.
  :issue:`3367` by :user:`Jeremy Steward <ThatGeoGuy>`

- Fixed a bug in :class:`manifold.TSNE` affecting convergence of the
  gradient descent. :issue:`8768` by :user:`David DeTomaso <deto>`.

- Fixed a bug in :class:`manifold.TSNE` where it stored the incorrect
  ``kl_divergence_``. :issue:`6507` by :user:`Sebastian Saeger <ssaeger>`.

- Fixed improper scaling in :class:`cross_decomposition.PLSRegression`
  with ``scale=True``. :issue:`7819` by :user:`jayzed82 <jayzed82>`.

- :class:`cluster.bicluster.SpectralCoclustering` and
  :class:`cluster.bicluster.SpectralBiclustering` ``fit`` method conforms
  with API by accepting ``y`` and returning the object.  :issue:`6126`,
  :issue:`7814` by :user:`Laurent Direr <ldirer>` and :user:`Maniteja
  Nandana <maniteja123>`.

- Fix bug where :mod:`mixture` ``sample`` methods did not return as many
  samples as requested. :issue:`7702` by :user:`Levi John Wolf <ljwolf>`.

- Fixed the shrinkage implementation in :class:`neighbors.NearestCentroid`.
  :issue:`9219` by `Hanmin Qin <https://github.com/qinhanmin2014>`_.

Preprocessing and feature selection

- For sparse matrices, :func:`preprocessing.normalize` with ``return_norm=True``
  will now raise a ``NotImplementedError`` with 'l1' or 'l2' norm and with
  norm 'max' the norms returned will be the same as for dense matrices.
  :issue:`7771` by `Ang Lu <https://github.com/luang008>`_.

- Fix a bug where :class:`feature_selection.SelectFdr` did not
  exactly implement Benjamini-Hochberg procedure. It formerly may have
  selected fewer features than it should.
  :issue:`7490` by :user:`Peng Meng <mpjlu>`.

- Fixed a bug where :class:`linear_model.RandomizedLasso` and
  :class:`linear_model.RandomizedLogisticRegression` breaks for
  sparse input. :issue:`8259` by :user:`Aman Dalmia <dalmia>`.

- Fix a bug where :class:`feature_extraction.FeatureHasher`
  mandatorily applied a sparse random projection to the hashed features,
  preventing the use of
  :class:`feature_extraction.text.HashingVectorizer` in a
  pipeline with  :class:`feature_extraction.text.TfidfTransformer`.
  :issue:`7565` by :user:`Roman Yurchak <rth>`.

- Fix a bug where :class:`feature_selection.mutual_info_regression` did not
  correctly use ``n_neighbors``. :issue:`8181` by :user:`Guillaume Lemaitre
  <glemaitre>`.

Model evaluation and meta-estimators

- Fixed a bug where :func:`model_selection.BaseSearchCV.inverse_transform`
  returns ``self.best_estimator_.transform()`` instead of
  ``self.best_estimator_.inverse_transform()``.
  :issue:`8344` by :user:`Akshay Gupta <Akshay0724>` and :user:`Rasmus Eriksson <MrMjauh>`.

- Added ``classes_`` attribute to :class:`model_selection.GridSearchCV`,
  :class:`model_selection.RandomizedSearchCV`,  :class:`grid_search.GridSearchCV`,
  and  :class:`grid_search.RandomizedSearchCV` that matches the ``classes_``
  attribute of ``best_estimator_``. :issue:`7661` and :issue:`8295`
  by :user:`Alyssa Batula <abatula>`, :user:`Dylan Werner-Meier <unautre>`,
  and :user:`Stephen Hoover <stephen-hoover>`.

- Fixed a bug where :func:`model_selection.validation_curve`
  reused the same estimator for each parameter value.
  :issue:`7365` by :user:`Aleksandr Sandrovskii <Sundrique>`.

- :func:`model_selection.permutation_test_score` now works with Pandas
  types. :issue:`5697` by :user:`Stijn Tonk <equialgo>`.

- Several fixes to input validation in
  :class:`multiclass.OutputCodeClassifier`
  :issue:`8086` by `Andreas Müller`_.

- :class:`multiclass.OneVsOneClassifier`'s ``partial_fit`` now ensures all
  classes are provided up-front. :issue:`6250` by
  :user:`Asish Panda <kaichogami>`.

- Fix :func:`multioutput.MultiOutputClassifier.predict_proba` to return a
  list of 2d arrays, rather than a 3d array. In the case where different
  target columns had different numbers of classes, a ``ValueError`` would be
  raised on trying to stack matrices with different dimensions.
  :issue:`8093` by :user:`Peter Bull <pjbull>`.

- Cross validation now works with Pandas datatypes that that have a
  read-only index. :issue:`9507` by `Loic Esteve`_.

Metrics

- :func:`metrics.average_precision_score` no longer linearly
  interpolates between operating points, and instead weighs precisions
  by the change in recall since the last operating point, as per the
  `Wikipedia entry <http://en.wikipedia.org/wiki/Average_precision>`_.
  (`#7356 <https://github.com/scikit-learn/scikit-learn/pull/7356>`_). By
  :user:`Nick Dingwall <ndingwall>` and `Gael Varoquaux`_.

- Fix a bug in :func:`metrics.classification._check_targets`
  which would return ``'binary'`` if ``y_true`` and ``y_pred`` were
  both ``'binary'`` but the union of ``y_true`` and ``y_pred`` was
  ``'multiclass'``. :issue:`8377` by `Loic Esteve`_.

- Fixed an integer overflow bug in :func:`metrics.confusion_matrix` and
  hence :func:`metrics.cohen_kappa_score`. :issue:`8354`, :issue:`7929`
  by `Joel Nothman`_ and :user:`Jon Crall <Erotemic>`.

- Fixed passing of ``gamma`` parameter to the ``chi2`` kernel in
  :func:`metrics.pairwise.pairwise_kernels` :issue:`5211` by
  :user:`Nick Rhinehart <nrhine1>`,
  :user:`Saurabh Bansod <mth4saurabh>` and `Andreas Müller`_.

Miscellaneous

- Fixed a bug when :func:`datasets.make_classification` fails
  when generating more than 30 features. :issue:`8159` by
  :user:`Herilalaina Rakotoarison <herilalaina>`.

- Fixed a bug where :func:`datasets.make_moons` gives an
  incorrect result when ``n_samples`` is odd.
  :issue:`8198` by :user:`Josh Levy <levy5674>`.

- Some ``fetch_`` functions in :mod:`datasets` were ignoring the
  ``download_if_missing`` keyword. :issue:`7944` by :user:`Ralf Gommers <rgommers>`.

- Fix estimators to accept a ``sample_weight`` parameter of type
  ``pandas.Series`` in their ``fit`` function. :issue:`7825` by
  `Kathleen Chen`_.

- Fix a bug in cases where ``numpy.cumsum`` may be numerically unstable,
  raising an exception if instability is identified. :issue:`7376` and
  :issue:`7331` by `Joel Nothman`_ and :user:`yangarbiter`.

- Fix a bug where :meth:`base.BaseEstimator.__getstate__`
  obstructed pickling customizations of child-classes, when used in a
  multiple inheritance context.
  :issue:`8316` by :user:`Holger Peters <HolgerPeters>`.

- Update Sphinx-Gallery from 0.1.4 to 0.1.7 for resolving links in
  documentation build with Sphinx>1.5 :issue:`8010`, :issue:`7986` by
  :user:`Oscar Najera <Titan-C>`

- Add ``data_home`` parameter to :func:`sklearn.datasets.fetch_kddcup99`.
  :issue:`9289` by `Loic Esteve`_.

- Fix dataset loaders using Python 3 version of makedirs to also work in
  Python 2. :issue:`9284` by :user:`Sebastin Santy <SebastinSanty>`.

- Several minor issues were fixed with thanks to the alerts of
  [lgtm.com](http://lgtm.com). :issue:`9278` by :user:`Jean Helie <jhelie>`,
  among others.

API changes summary
-------------------

Trees and ensembles

- Gradient boosting base models are no longer estimators. By `Andreas Müller`_.

- All tree based estimators now accept a ``min_impurity_decrease``
  parameter in lieu of the ``min_impurity_split``, which is now deprecated.
  The ``min_impurity_decrease`` helps stop splitting the nodes in which
  the weighted impurity decrease from splitting is no longer alteast
  ``min_impurity_decrease``.  :issue:`8449` by `Raghav RV`_.

Linear, kernelized and related models

- ``n_iter`` parameter is deprecated in :class:`linear_model.SGDClassifier`,
  :class:`linear_model.SGDRegressor`,
  :class:`linear_model.PassiveAggressiveClassifier`,
  :class:`linear_model.PassiveAggressiveRegressor` and
  :class:`linear_model.Perceptron`. By `Tom Dupre la Tour`_.

Other predictors

- :class:`neighbors.LSHForest` has been deprecated and will be
  removed in 0.21 due to poor performance.
  :issue:`9078` by :user:`Laurent Direr <ldirer>`.

- :class:`neighbors.NearestCentroid` no longer purports to support
  ``metric='precomputed'`` which now raises an error. :issue:`8515` by
  :user:`Sergul Aydore <sergulaydore>`.

- The ``alpha`` parameter of :class:`semi_supervised.LabelPropagation` now
  has no effect and is deprecated to be removed in 0.21. :issue:`9239`
  by :user:`Andre Ambrosio Boechat <boechat107>`, :user:`Utkarsh Upadhyay
  <musically-ut>`, and `Joel Nothman`_.

Decomposition, manifold learning and clustering

- Deprecate the ``doc_topic_distr`` argument of the ``perplexity`` method
  in :class:`decomposition.LatentDirichletAllocation` because the
  user no longer has access to the unnormalized document topic distribution
  needed for the perplexity calculation. :issue:`7954` by
  :user:`Gary Foreman <garyForeman>`.

- The ``n_topics`` parameter of :class:`decomposition.LatentDirichletAllocation`
  has been renamed to ``n_components`` and will be removed in version 0.21.
  :issue:`8922` by :user:`Attractadore`.

- :meth:`decomposition.SparsePCA.transform`'s ``ridge_alpha`` parameter is
  deprecated in preference for class parameter.
  :issue:`8137` by :user:`Naoya Kanai <naoyak>`.

- :class:`cluster.DBSCAN` now has a ``metric_params`` parameter.
  :issue:`8139` by :user:`Naoya Kanai <naoyak>`.

Preprocessing and feature selection

- :class:`feature_selection.SelectFromModel` now has a ``partial_fit``
  method only if the underlying estimator does. By `Andreas Müller`_.

- :class:`feature_selection.SelectFromModel` now validates the ``threshold``
  parameter and sets the ``threshold_`` attribute during the call to
  ``fit``, and no longer during the call to ``transform```. By `Andreas
  Müller`_.

- The ``non_negative`` parameter in :class:`feature_extraction.FeatureHasher`
  has been deprecated, and replaced with a more principled alternative,
  ``alternate_sign``.
  :issue:`7565` by :user:`Roman Yurchak <rth>`.

- :class:`linear_model.RandomizedLogisticRegression`,
  and :class:`linear_model.RandomizedLasso` have been deprecated and will
  be removed in version 0.21.
  :issue:`8995` by :user:`Ramana.S <sentient07>`.

Model evaluation and meta-estimators

- Deprecate the ``fit_params`` constructor input to the
  :class:`model_selection.GridSearchCV` and
  :class:`model_selection.RandomizedSearchCV` in favor
  of passing keyword parameters to the ``fit`` methods
  of those classes. Data-dependent parameters needed for model
  training should be passed as keyword arguments to ``fit``,
  and conforming to this convention will allow the hyperparameter
  selection classes to be used with tools such as
  :func:`model_selection.cross_val_predict`.
  :issue:`2879` by :user:`Stephen Hoover <stephen-hoover>`.

- In version 0.21, the default behavior of splitters that use the
  ``test_size`` and ``train_size`` parameter will change, such that
  specifying ``train_size`` alone will cause ``test_size`` to be the
  remainder. :issue:`7459` by :user:`Nelson Liu <nelson-liu>`.

- :class:`multiclass.OneVsRestClassifier` now has ``partial_fit``,
  ``decision_function`` and ``predict_proba`` methods only when the
  underlying estimator does.  :issue:`7812` by `Andreas Müller`_ and
  :user:`Mikhail Korobov <kmike>`.

- :class:`multiclass.OneVsRestClassifier` now has a ``partial_fit`` method
  only if the underlying estimator does.  By `Andreas Müller`_.

- The ``decision_function`` output shape for binary classification in
  :class:`multiclass.OneVsRestClassifier` and
  :class:`multiclass.OneVsOneClassifier` is now ``(n_samples,)`` to conform
  to scikit-learn conventions. :issue:`9100` by `Andreas Müller`_.

- The :func:`multioutput.MultiOutputClassifier.predict_proba`
  function used to return a 3d array (``n_samples``, ``n_classes``,
  ``n_outputs``). In the case where different target columns had different
  numbers of classes, a ``ValueError`` would be raised on trying to stack
  matrices with different dimensions. This function now returns a list of
  arrays where the length of the list is ``n_outputs``, and each array is
  (``n_samples``, ``n_classes``) for that particular output.
  :issue:`8093` by :user:`Peter Bull <pjbull>`.

- Replace attribute ``named_steps`` ``dict`` to :class:`utils.Bunch`
  in :class:`pipeline.Pipeline` to enable tab completion in interactive
  environment. In the case conflict value on ``named_steps`` and ``dict``
  attribute, ``dict`` behavior will be prioritized.
  :issue:`8481` by :user:`Herilalaina Rakotoarison <herilalaina>`.

Miscellaneous

- Deprecate the ``y`` parameter in ``transform`` and ``inverse_transform``.
  The method  should not accept ``y`` parameter, as it's used at the prediction time.
  :issue:`8174` by :user:`Tahar Zanouda <tzano>`, `Alexandre Gramfort`_
  and `Raghav RV`_.

- SciPy >= 0.13.3 and NumPy >= 1.8.2 are now the minimum supported versions
  for scikit-learn. The following backported functions in
  :mod:`utils` have been removed or deprecated accordingly.
  :issue:`8854` and :issue:`8874` by :user:`Naoya Kanai <naoyak>`

- The ``store_covariances`` and ``covariances_`` parameters of
  :class:`discriminant_analysis.QuadraticDiscriminantAnalysis`
  has been renamed to ``store_covariance`` and ``covariance_`` to be
  consistent with the corresponding parameter names of the
  :class:`discriminant_analysis.LinearDiscriminantAnalysis`. They will be
  removed in version 0.21. :issue:`7998` by :user:`Jiacheng <mrbeann>`

  Removed in 0.19:

  - ``utils.fixes.argpartition``
  - ``utils.fixes.array_equal``
  - ``utils.fixes.astype``
  - ``utils.fixes.bincount``
  - ``utils.fixes.expit``
  - ``utils.fixes.frombuffer_empty``
  - ``utils.fixes.in1d``
  - ``utils.fixes.norm``
  - ``utils.fixes.rankdata``
  - ``utils.fixes.safe_copy``

  Deprecated in 0.19, to be removed in 0.21:

  - ``utils.arpack.eigs``
  - ``utils.arpack.eigsh``
  - ``utils.arpack.svds``
  - ``utils.extmath.fast_dot``
  - ``utils.extmath.logsumexp``
  - ``utils.extmath.norm``
  - ``utils.extmath.pinvh``
  - ``utils.graph.graph_laplacian``
  - ``utils.random.choice``
  - ``utils.sparsetools.connected_components``
  - ``utils.stats.rankdata``

- Estimators with both methods ``decision_function`` and ``predict_proba``
  are now required to have a monotonic relation between them. The
  method ``check_decision_proba_consistency`` has been added in
  **utils.estimator_checks** to check their consistency.
  :issue:`7578` by :user:`Shubham Bhardwaj <shubham0704>`

- All checks in ``utils.estimator_checks``, in particular
  :func:`utils.estimator_checks.check_estimator` now accept estimator
  instances. Most other checks do not accept
  estimator classes any more. :issue:`9019` by `Andreas Müller`_.

- Ensure that estimators' attributes ending with ``_`` are not set
  in the constructor but only in the ``fit`` method. Most notably,
  ensemble estimators (deriving from :class:`ensemble.BaseEnsemble`)
  now only have ``self.estimators_`` available after ``fit``.
  :issue:`7464` by `Lars Buitinck`_ and `Loic Esteve`_.


Code and Documentation Contributors
-----------------------------------

Thanks to everyone who has contributed to the maintenance and improvement of the
project since version 0.18, including:

Joel Nothman, Loic Esteve, Andreas Mueller, Guillaume Lemaitre, Olivier Grisel,
Hanmin Qin, Raghav RV, Alexandre Gramfort, themrmax, Aman Dalmia, Gael
Varoquaux, Naoya Kanai, Tom Dupré la Tour, Rishikesh, Nelson Liu, Taehoon Lee,
Nelle Varoquaux, Aashil, Mikhail Korobov, Sebastin Santy, Joan Massich, Roman
Yurchak, RAKOTOARISON Herilalaina, Thierry Guillemot, Alexandre Abadie, Carol
Willing, Balakumaran Manoharan, Josh Karnofsky, Vlad Niculae, Utkarsh Upadhyay,
Dmitry Petrov, Minghui Liu, Srivatsan, Vincent Pham, Albert Thomas, Jake
VanderPlas, Attractadore, JC Liu, alexandercbooth, chkoar, Óscar Nájera,
Aarshay Jain, Kyle Gilliam, Ramana Subramanyam, CJ Carey, Clement Joudet, David
Robles, He Chen, Joris Van den Bossche, Karan Desai, Katie Luangkote, Leland
McInnes, Maniteja Nandana, Michele Lacchia, Sergei Lebedev, Shubham Bhardwaj,
akshay0724, omtcyfz, rickiepark, waterponey, Vathsala Achar, jbDelafosse, Ralf
Gommers, Ekaterina Krivich, Vivek Kumar, Ishank Gulati, Dave Elliott, ldirer,
Reiichiro Nakano, Levi John Wolf, Mathieu Blondel, Sid Kapur, Dougal J.
Sutherland, midinas, mikebenfield, Sourav Singh, Aseem Bansal, Ibraim Ganiev,
Stephen Hoover, AishwaryaRK, Steven C. Howell, Gary Foreman, Neeraj Gangwar,
Tahar, Jon Crall, dokato, Kathy Chen, ferria, Thomas Moreau, Charlie Brummitt,
Nicolas Goix, Adam Kleczewski, Sam Shleifer, Nikita Singh, Basil Beirouti,
Giorgio Patrini, Manoj Kumar, Rafael Possas, James Bourbeau, James A. Bednar,
Janine Harper, Jaye, Jean Helie, Jeremy Steward, Artsiom, John Wei, Jonathan
LIgo, Jonathan Rahn, seanpwilliams, Arthur Mensch, Josh Levy, Julian Kuhlmann,
Julien Aubert, Jörn Hees, Kai, shivamgargsya, Kat Hempstalk, Kaushik
Lakshmikanth, Kennedy, Kenneth Lyons, Kenneth Myers, Kevin Yap, Kirill Bobyrev,
Konstantin Podshumok, Arthur Imbert, Lee Murray, toastedcornflakes, Lera, Li
Li, Arthur Douillard, Mainak Jas, tobycheese, Manraj Singh, Manvendra Singh,
Marc Meketon, MarcoFalke, Matthew Brett, Matthias Gilch, Mehul Ahuja, Melanie
Goetz, Meng, Peng, Michael Dezube, Michal Baumgartner, vibrantabhi19, Artem
Golubin, Milen Paskov, Antonin Carette, Morikko, MrMjauh, NALEPA Emmanuel,
Namiya, Antoine Wendlinger, Narine Kokhlikyan, NarineK, Nate Guerin, Angus
Williams, Ang Lu, Nicole Vavrova, Nitish Pandey, Okhlopkov Daniil Olegovich,
Andy Craze, Om Prakash, Parminder Singh, Patrick Carlson, Patrick Pei, Paul
Ganssle, Paulo Haddad, Paweł Lorek, Peng Yu, Pete Bachant, Peter Bull, Peter
Csizsek, Peter Wang, Pieter Arthur de Jong, Ping-Yao, Chang, Preston Parry,
Puneet Mathur, Quentin Hibon, Andrew Smith, Andrew Jackson, 1kastner, Rameshwar
Bhaskaran, Rebecca Bilbro, Remi Rampin, Andrea Esuli, Rob Hall, Robert
Bradshaw, Romain Brault, Aman Pratik, Ruifeng Zheng, Russell Smith, Sachin
Agarwal, Sailesh Choyal, Samson Tan, Samuël Weber, Sarah Brown, Sebastian
Pölsterl, Sebastian Raschka, Sebastian Saeger, Alyssa Batula, Abhyuday Pratap
Singh, Sergey Feldman, Sergul Aydore, Sharan Yalburgi, willduan, Siddharth
Gupta, Sri Krishna, Almer, Stijn Tonk, Allen Riddell, Theofilos Papapanagiotou,
Alison, Alexis Mignon, Tommy Boucher, Tommy Löfstedt, Toshihiro Kamishima,
Tyler Folkman, Tyler Lanigan, Alexander Junge, Varun Shenoy, Victor Poughon,
Vilhelm von Ehrenheim, Aleksandr Sandrovskii, Alan Yee, Vlasios Vasileiou,
Warut Vijitbenjaronk, Yang Zhang, Yaroslav Halchenko, Yichuan Liu, Yuichi
Fujikawa, affanv14, aivision2020, xor, andreh7, brady salz, campustrampus,
Agamemnon Krasoulis, ditenberg, elena-sharova, filipj8, fukatani, gedeck,
guiniol, guoci, hakaa1, hongkahjun, i-am-xhy, jakirkham, jaroslaw-weber,
jayzed82, jeroko, jmontoyam, jonathan.striebel, josephsalmon, jschendel,
leereeves, martin-hahn, mathurinm, mehak-sachdeva, mlewis1729, mlliou112,
mthorrell, ndingwall, nuffe, yangarbiter, plagree, pldtc325, Breno Freitas,
Brett Olsen, Brian A. Alfano, Brian Burns, polmauri, Brandon Carter, Charlton
Austin, Chayant T15h, Chinmaya Pancholi, Christian Danielsen, Chung Yen,
Chyi-Kwei Yau, pravarmahajan, DOHMATOB Elvis, Daniel LeJeune, Daniel Hnyk,
Darius Morawiec, David DeTomaso, David Gasquez, David Haberthür, David
Heryanto, David Kirkby, David Nicholson, rashchedrin, Deborah Gertrude Digges,
Denis Engemann, Devansh D, Dickson, Bob Baxley, Don86, E. Lynch-Klarup, Ed
Rogers, Elizabeth Ferriss, Ellen-Co2, Fabian Egli, Fang-Chieh Chou, Bing Tian
Dai, Greg Stupp, Grzegorz Szpak, Bertrand Thirion, Hadrien Bertrand, Harizo
Rajaona, zxcvbnius, Henry Lin, Holger Peters, Icyblade Dai, Igor
Andriushchenko, Ilya, Isaac Laughlin, Iván Vallés, Aurélien Bellet, JPFrancoia,
Jacob Schreiber, Asish Mahapatra

.. _changes_0_18_2:

Version 0.18.2
==============

**June 20, 2017**

.. topic:: Last release with Python 2.6 support

    Scikit-learn 0.18 is the last major release of scikit-learn to support Python 2.6.
    Later versions of scikit-learn will require Python 2.7 or above.


Changelog
---------

- Fixes for compatibility with NumPy 1.13.0: :issue:`7946` :issue:`8355` by
  `Loic Esteve`_.

- Minor compatibility changes in the examples :issue:`9010` :issue:`8040`
  :issue:`9149`.

Code Contributors
-----------------
Aman Dalmia, Loic Esteve, Nate Guerin, Sergei Lebedev


.. _changes_0_18_1:

Version 0.18.1
==============

**November 11, 2016**

Changelog
---------

Enhancements
............

- Improved ``sample_without_replacement`` speed by utilizing
  numpy.random.permutation for most cases. As a result,
  samples may differ in this release for a fixed random state.
  Affected estimators:

  - :class:`ensemble.BaggingClassifier`
  - :class:`ensemble.BaggingRegressor`
  - :class:`linear_model.RANSACRegressor`
  - :class:`model_selection.RandomizedSearchCV`
  - :class:`random_projection.SparseRandomProjection`

  This also affects the :meth:`datasets.make_classification`
  method.

Bug fixes
.........

- Fix issue where ``min_grad_norm`` and ``n_iter_without_progress``
  parameters were not being utilised by :class:`manifold.TSNE`.
  :issue:`6497` by :user:`Sebastian Säger <ssaeger>`

- Fix bug for svm's decision values when ``decision_function_shape``
  is ``ovr`` in :class:`svm.SVC`.
  :class:`svm.SVC`'s decision_function was incorrect from versions
  0.17.0 through 0.18.0.
  :issue:`7724` by `Bing Tian Dai`_

- Attribute ``explained_variance_ratio`` of
  :class:`discriminant_analysis.LinearDiscriminantAnalysis` calculated
  with SVD and Eigen solver are now of the same length. :issue:`7632`
  by :user:`JPFrancoia <JPFrancoia>`

- Fixes issue in :ref:`univariate_feature_selection` where score
  functions were not accepting multi-label targets. :issue:`7676`
  by :user:`Mohammed Affan <affanv14>`

- Fixed setting parameters when calling ``fit`` multiple times on
  :class:`feature_selection.SelectFromModel`. :issue:`7756` by `Andreas Müller`_

- Fixes issue in ``partial_fit`` method of
  :class:`multiclass.OneVsRestClassifier` when number of classes used in
  ``partial_fit`` was less than the total number of classes in the
  data. :issue:`7786` by `Srivatsan Ramesh`_

- Fixes issue in :class:`calibration.CalibratedClassifierCV` where
  the sum of probabilities of each class for a data was not 1, and
  ``CalibratedClassifierCV`` now handles the case where the training set
  has less number of classes than the total data. :issue:`7799` by
  `Srivatsan Ramesh`_

- Fix a bug where :class:`sklearn.feature_selection.SelectFdr` did not
  exactly implement Benjamini-Hochberg procedure. It formerly may have
  selected fewer features than it should.
  :issue:`7490` by :user:`Peng Meng <mpjlu>`.

- :class:`sklearn.manifold.LocallyLinearEmbedding` now correctly handles
  integer inputs. :issue:`6282` by `Jake Vanderplas`_.

- The ``min_weight_fraction_leaf`` parameter of tree-based classifiers and
  regressors now assumes uniform sample weights by default if the
  ``sample_weight`` argument is not passed to the ``fit`` function.
  Previously, the parameter was silently ignored. :issue:`7301`
  by :user:`Nelson Liu <nelson-liu>`.

- Numerical issue with :class:`linear_model.RidgeCV` on centered data when
  `n_features > n_samples`. :issue:`6178` by `Bertrand Thirion`_

- Tree splitting criterion classes' cloning/pickling is now memory safe
  :issue:`7680` by :user:`Ibraim Ganiev <olologin>`.

- Fixed a bug where :class:`decomposition.NMF` sets its ``n_iters_``
  attribute in `transform()`. :issue:`7553` by :user:`Ekaterina
  Krivich <kiote>`.

- :class:`sklearn.linear_model.LogisticRegressionCV` now correctly handles
  string labels. :issue:`5874` by `Raghav RV`_.

- Fixed a bug where :func:`sklearn.model_selection.train_test_split` raised
  an error when ``stratify`` is a list of string labels. :issue:`7593` by
  `Raghav RV`_.

- Fixed a bug where :class:`sklearn.model_selection.GridSearchCV` and
  :class:`sklearn.model_selection.RandomizedSearchCV` were not pickleable
  because of a pickling bug in ``np.ma.MaskedArray``. :issue:`7594` by
  `Raghav RV`_.

- All cross-validation utilities in :mod:`sklearn.model_selection` now
  permit one time cross-validation splitters for the ``cv`` parameter. Also
  non-deterministic cross-validation splitters (where multiple calls to
  ``split`` produce dissimilar splits) can be used as ``cv`` parameter.
  The :class:`sklearn.model_selection.GridSearchCV` will cross-validate each
  parameter setting on the split produced by the first ``split`` call
  to the cross-validation splitter.  :issue:`7660` by `Raghav RV`_.

- Fix bug where :meth:`preprocessing.MultiLabelBinarizer.fit_transform`
  returned an invalid CSR matrix.
  :issue:`7750` by :user:`CJ Carey <perimosocordiae>`.

- Fixed a bug where :func:`metrics.pairwise.cosine_distances` could return a
  small negative distance. :issue:`7732` by :user:`Artsion <asanakoy>`.

API changes summary
-------------------

Trees and forests

- The ``min_weight_fraction_leaf`` parameter of tree-based classifiers and
  regressors now assumes uniform sample weights by default if the
  ``sample_weight`` argument is not passed to the ``fit`` function.
  Previously, the parameter was silently ignored. :issue:`7301` by :user:`Nelson
  Liu <nelson-liu>`.

- Tree splitting criterion classes' cloning/pickling is now memory safe.
  :issue:`7680` by :user:`Ibraim Ganiev <olologin>`.


Linear, kernelized and related models

- Length of ``explained_variance_ratio`` of
  :class:`discriminant_analysis.LinearDiscriminantAnalysis`
  changed for both Eigen and SVD solvers. The attribute has now a length
  of min(n_components, n_classes - 1). :issue:`7632`
  by :user:`JPFrancoia <JPFrancoia>`

- Numerical issue with :class:`linear_model.RidgeCV` on centered data when
  ``n_features > n_samples``. :issue:`6178` by `Bertrand Thirion`_

.. _changes_0_18:

Version 0.18
============

**September 28, 2016**

.. topic:: Last release with Python 2.6 support

    Scikit-learn 0.18 will be the last version of scikit-learn to support Python 2.6.
    Later versions of scikit-learn will require Python 2.7 or above.

.. _model_selection_changes:

Model Selection Enhancements and API Changes
--------------------------------------------

- **The model_selection module**

  The new module :mod:`sklearn.model_selection`, which groups together the
  functionalities of formerly :mod:`sklearn.cross_validation`,
  :mod:`sklearn.grid_search` and :mod:`sklearn.learning_curve`, introduces new
  possibilities such as nested cross-validation and better manipulation of
  parameter searches with Pandas.

  Many things will stay the same but there are some key differences. Read
  below to know more about the changes.

- **Data-independent CV splitters enabling nested cross-validation**

  The new cross-validation splitters, defined in the
  :mod:`sklearn.model_selection`, are no longer initialized with any
  data-dependent parameters such as ``y``. Instead they expose a
  :func:`split` method that takes in the data and yields a generator for the
  different splits.

  This change makes it possible to use the cross-validation splitters to
  perform nested cross-validation, facilitated by
  :class:`model_selection.GridSearchCV` and
  :class:`model_selection.RandomizedSearchCV` utilities.

- **The enhanced cv_results_ attribute**

  The new ``cv_results_`` attribute (of :class:`model_selection.GridSearchCV`
  and :class:`model_selection.RandomizedSearchCV`) introduced in lieu of the
  ``grid_scores_`` attribute is a dict of 1D arrays with elements in each
  array corresponding to the parameter settings (i.e. search candidates).

  The ``cv_results_`` dict can be easily imported into ``pandas`` as a
  ``DataFrame`` for exploring the search results.

  The ``cv_results_`` arrays include scores for each cross-validation split
  (with keys such as ``'split0_test_score'``), as well as their mean
  (``'mean_test_score'``) and standard deviation (``'std_test_score'``).

  The ranks for the search candidates (based on their mean
  cross-validation score) is available at ``cv_results_['rank_test_score']``.

  The parameter values for each parameter is stored separately as numpy
  masked object arrays. The value, for that search candidate, is masked if
  the corresponding parameter is not applicable. Additionally a list of all
  the parameter dicts are stored at ``cv_results_['params']``.

- **Parameters n_folds and n_iter renamed to n_splits**

  Some parameter names have changed:
  The ``n_folds`` parameter in new :class:`model_selection.KFold`,
  :class:`model_selection.GroupKFold` (see below for the name change),
  and :class:`model_selection.StratifiedKFold` is now renamed to
  ``n_splits``. The ``n_iter`` parameter in
  :class:`model_selection.ShuffleSplit`, the new class
  :class:`model_selection.GroupShuffleSplit` and
  :class:`model_selection.StratifiedShuffleSplit` is now renamed to
  ``n_splits``.

- **Rename of splitter classes which accepts group labels along with data**

  The cross-validation splitters ``LabelKFold``,
  ``LabelShuffleSplit``, ``LeaveOneLabelOut`` and ``LeavePLabelOut`` have
  been renamed to :class:`model_selection.GroupKFold`,
  :class:`model_selection.GroupShuffleSplit`,
  :class:`model_selection.LeaveOneGroupOut` and
  :class:`model_selection.LeavePGroupsOut` respectively.

  Note the change from singular to plural form in
  :class:`model_selection.LeavePGroupsOut`.

- **Fit parameter labels renamed to groups**

  The ``labels`` parameter in the :func:`split` method of the newly renamed
  splitters :class:`model_selection.GroupKFold`,
  :class:`model_selection.LeaveOneGroupOut`,
  :class:`model_selection.LeavePGroupsOut`,
  :class:`model_selection.GroupShuffleSplit` is renamed to ``groups``
  following the new nomenclature of their class names.

- **Parameter n_labels renamed to n_groups**

  The parameter ``n_labels`` in the newly renamed
  :class:`model_selection.LeavePGroupsOut` is changed to ``n_groups``.

- Training scores and Timing information

  ``cv_results_`` also includes the training scores for each
  cross-validation split (with keys such as ``'split0_train_score'``), as
  well as their mean (``'mean_train_score'``) and standard deviation
  (``'std_train_score'``). To avoid the cost of evaluating training score,
  set ``return_train_score=False``.

  Additionally the mean and standard deviation of the times taken to split,
  train and score the model across all the cross-validation splits is
  available at the key ``'mean_time'`` and ``'std_time'`` respectively.

Changelog
---------

New features
............

Classifiers and Regressors

- The Gaussian Process module has been reimplemented and now offers classification
  and regression estimators through :class:`gaussian_process.GaussianProcessClassifier`
  and  :class:`gaussian_process.GaussianProcessRegressor`. Among other things, the new
  implementation supports kernel engineering, gradient-based hyperparameter optimization or
  sampling of functions from GP prior and GP posterior. Extensive documentation and
  examples are provided. By `Jan Hendrik Metzen`_.

- Added new supervised learning algorithm: :ref:`Multi-layer Perceptron <multilayer_perceptron>`
  :issue:`3204` by :user:`Issam H. Laradji <IssamLaradji>`

- Added :class:`linear_model.HuberRegressor`, a linear model robust to outliers.
  :issue:`5291` by `Manoj Kumar`_.

- Added the :class:`multioutput.MultiOutputRegressor` meta-estimator. It
  converts single output regressors to multi-output regressors by fitting
  one regressor per output. By :user:`Tim Head <betatim>`.

Other estimators

- New :class:`mixture.GaussianMixture` and :class:`mixture.BayesianGaussianMixture`
  replace former mixture models, employing faster inference
  for sounder results. :issue:`7295` by :user:`Wei Xue <xuewei4d>` and
  :user:`Thierry Guillemot <tguillemot>`.

- Class :class:`decomposition.RandomizedPCA` is now factored into :class:`decomposition.PCA`
  and it is available calling with parameter ``svd_solver='randomized'``.
  The default number of ``n_iter`` for ``'randomized'`` has changed to 4. The old
  behavior of PCA is recovered by ``svd_solver='full'``. An additional solver
  calls ``arpack`` and performs truncated (non-randomized) SVD. By default,
  the best solver is selected depending on the size of the input and the
  number of components requested. :issue:`5299` by :user:`Giorgio Patrini <giorgiop>`.

- Added two functions for mutual information estimation:
  :func:`feature_selection.mutual_info_classif` and
  :func:`feature_selection.mutual_info_regression`. These functions can be
  used in :class:`feature_selection.SelectKBest` and
  :class:`feature_selection.SelectPercentile` as score functions.
  By :user:`Andrea Bravi <AndreaBravi>` and :user:`Nikolay Mayorov <nmayorov>`.

- Added the :class:`ensemble.IsolationForest` class for anomaly detection based on
  random forests. By `Nicolas Goix`_.

- Added ``algorithm="elkan"`` to :class:`cluster.KMeans` implementing
  Elkan's fast K-Means algorithm. By `Andreas Müller`_.

Model selection and evaluation

- Added :func:`metrics.cluster.fowlkes_mallows_score`, the Fowlkes Mallows
  Index which measures the similarity of two clusterings of a set of points
  By :user:`Arnaud Fouchet <afouchet>` and :user:`Thierry Guillemot <tguillemot>`.

- Added :func:`metrics.calinski_harabaz_score`, which computes the Calinski
  and Harabaz score to evaluate the resulting clustering of a set of points.
  By :user:`Arnaud Fouchet <afouchet>` and :user:`Thierry Guillemot <tguillemot>`.

- Added new cross-validation splitter
  :class:`model_selection.TimeSeriesSplit` to handle time series data.
  :issue:`6586` by :user:`YenChen Lin <yenchenlin>`

- The cross-validation iterators are replaced by cross-validation splitters
  available from :mod:`sklearn.model_selection`, allowing for nested
  cross-validation. See :ref:`model_selection_changes` for more information.
  :issue:`4294` by `Raghav RV`_.

Enhancements
............

Trees and ensembles

- Added a new splitting criterion for :class:`tree.DecisionTreeRegressor`,
  the mean absolute error. This criterion can also be used in
  :class:`ensemble.ExtraTreesRegressor`,
  :class:`ensemble.RandomForestRegressor`, and the gradient boosting
  estimators. :issue:`6667` by :user:`Nelson Liu <nelson-liu>`.

- Added weighted impurity-based early stopping criterion for decision tree
  growth. :issue:`6954` by :user:`Nelson Liu <nelson-liu>`

- The random forest, extra tree and decision tree estimators now has a
  method ``decision_path`` which returns the decision path of samples in
  the tree. By `Arnaud Joly`_.

- A new example has been added unveiling the decision tree structure.
  By `Arnaud Joly`_.

- Random forest, extra trees, decision trees and gradient boosting estimator
  accept the parameter ``min_samples_split`` and ``min_samples_leaf``
  provided as a percentage of the training samples. By :user:`yelite <yelite>` and `Arnaud Joly`_.

- Gradient boosting estimators accept the parameter ``criterion`` to specify
  to splitting criterion used in built decision trees.
  :issue:`6667` by :user:`Nelson Liu <nelson-liu>`.

- The memory footprint is reduced (sometimes greatly) for
  :class:`ensemble.bagging.BaseBagging` and classes that inherit from it,
  i.e, :class:`ensemble.BaggingClassifier`,
  :class:`ensemble.BaggingRegressor`, and :class:`ensemble.IsolationForest`,
  by dynamically generating attribute ``estimators_samples_`` only when it is
  needed. By :user:`David Staub <staubda>`.

- Added ``n_jobs`` and ``sample_weight`` parameters for
  :class:`ensemble.VotingClassifier` to fit underlying estimators in parallel.
  :issue:`5805` by :user:`Ibraim Ganiev <olologin>`.

Linear, kernelized and related models

- In :class:`linear_model.LogisticRegression`, the SAG solver is now
  available in the multinomial case. :issue:`5251` by `Tom Dupre la Tour`_.

- :class:`linear_model.RANSACRegressor`, :class:`svm.LinearSVC` and
  :class:`svm.LinearSVR` now support ``sample_weight``.
  By :user:`Imaculate <Imaculate>`.

- Add parameter ``loss`` to :class:`linear_model.RANSACRegressor` to measure the
  error on the samples for every trial. By `Manoj Kumar`_.

- Prediction of out-of-sample events with Isotonic Regression
  (:class:`isotonic.IsotonicRegression`) is now much faster (over 1000x in tests with synthetic
  data). By :user:`Jonathan Arfa <jarfa>`.

- Isotonic regression (:class:`isotonic.IsotonicRegression`) now uses a better algorithm to avoid
  `O(n^2)` behavior in pathological cases, and is also generally faster
  (:issue:`#6691`). By `Antony Lee`_.

- :class:`naive_bayes.GaussianNB` now accepts data-independent class-priors
  through the parameter ``priors``. By :user:`Guillaume Lemaitre <glemaitre>`.

- :class:`linear_model.ElasticNet` and :class:`linear_model.Lasso`
  now works with ``np.float32`` input data without converting it
  into ``np.float64``. This allows to reduce the memory
  consumption. :issue:`6913` by :user:`YenChen Lin <yenchenlin>`.

- :class:`semi_supervised.LabelPropagation` and :class:`semi_supervised.LabelSpreading`
  now accept arbitrary kernel functions in addition to strings ``knn`` and ``rbf``.
  :issue:`5762` by :user:`Utkarsh Upadhyay <musically-ut>`.

Decomposition, manifold learning and clustering

- Added ``inverse_transform`` function to :class:`decomposition.NMF` to compute
  data matrix of original shape. By :user:`Anish Shah <AnishShah>`.

- :class:`cluster.KMeans` and :class:`cluster.MiniBatchKMeans` now works
  with ``np.float32`` and ``np.float64`` input data without converting it.
  This allows to reduce the memory consumption by using ``np.float32``.
  :issue:`6846` by :user:`Sebastian Säger <ssaeger>` and
  :user:`YenChen Lin <yenchenlin>`.

Preprocessing and feature selection

- :class:`preprocessing.RobustScaler` now accepts ``quantile_range`` parameter.
  :issue:`5929` by :user:`Konstantin Podshumok <podshumok>`.

- :class:`feature_extraction.FeatureHasher` now accepts string values.
  :issue:`6173` by :user:`Ryad Zenine <ryadzenine>` and
  :user:`Devashish Deshpande <dsquareindia>`.

- Keyword arguments can now be supplied to ``func`` in
  :class:`preprocessing.FunctionTransformer` by means of the ``kw_args``
  parameter. By `Brian McFee`_.

- :class:`feature_selection.SelectKBest` and :class:`feature_selection.SelectPercentile`
  now accept score functions that take X, y as input and return only the scores.
  By :user:`Nikolay Mayorov <nmayorov>`.

Model evaluation and meta-estimators

- :class:`multiclass.OneVsOneClassifier` and :class:`multiclass.OneVsRestClassifier`
  now support ``partial_fit``. By :user:`Asish Panda <kaichogami>` and
  :user:`Philipp Dowling <phdowling>`.

- Added support for substituting or disabling :class:`pipeline.Pipeline`
  and :class:`pipeline.FeatureUnion` components using the ``set_params``
  interface that powers :mod:`sklearn.grid_search`.
  See :ref:`sphx_glr_auto_examples_plot_compare_reduction.py`
  By `Joel Nothman`_ and :user:`Robert McGibbon <rmcgibbo>`.

- The new ``cv_results_`` attribute of :class:`model_selection.GridSearchCV`
  (and :class:`model_selection.RandomizedSearchCV`) can be easily imported
  into pandas as a ``DataFrame``. Ref :ref:`model_selection_changes` for
  more information. :issue:`6697` by `Raghav RV`_.

- Generalization of :func:`model_selection.cross_val_predict`.
  One can pass method names such as `predict_proba` to be used in the cross
  validation framework instead of the default `predict`.
  By :user:`Ori Ziv <zivori>` and :user:`Sears Merritt <merritts>`.

- The training scores and time taken for training followed by scoring for
  each search candidate are now available at the ``cv_results_`` dict.
  See :ref:`model_selection_changes` for more information.
  :issue:`7325` by :user:`Eugene Chen <eyc88>` and `Raghav RV`_.

Metrics

- Added ``labels`` flag to :class:`metrics.log_loss` to explicitly provide
  the labels when the number of classes in ``y_true`` and ``y_pred`` differ.
  :issue:`7239` by :user:`Hong Guangguo <hongguangguo>` with help from
  :user:`Mads Jensen <indianajensen>` and :user:`Nelson Liu <nelson-liu>`.

- Support sparse contingency matrices in cluster evaluation
  (:mod:`metrics.cluster.supervised`) to scale to a large number of
  clusters.
  :issue:`7419` by :user:`Gregory Stupp <stuppie>` and `Joel Nothman`_.

- Add ``sample_weight`` parameter to :func:`metrics.matthews_corrcoef`.
  By :user:`Jatin Shah <jatinshah>` and `Raghav RV`_.

- Speed up :func:`metrics.silhouette_score` by using vectorized operations.
  By `Manoj Kumar`_.

- Add ``sample_weight`` parameter to :func:`metrics.confusion_matrix`.
  By :user:`Bernardo Stein <DanielSidhion>`.

Miscellaneous

- Added ``n_jobs`` parameter to :class:`feature_selection.RFECV` to compute
  the score on the test folds in parallel. By `Manoj Kumar`_

- Codebase does not contain C/C++ cython generated files: they are
  generated during build. Distribution packages will still contain generated
  C/C++ files. By :user:`Arthur Mensch <arthurmensch>`.

- Reduce the memory usage for 32-bit float input arrays of
  :func:`utils.sparse_func.mean_variance_axis` and
  :func:`utils.sparse_func.incr_mean_variance_axis` by supporting cython
  fused types. By :user:`YenChen Lin <yenchenlin>`.

- The :func:`ignore_warnings` now accept a category argument to ignore only
  the warnings of a specified type. By :user:`Thierry Guillemot <tguillemot>`.

- Added parameter ``return_X_y`` and return type ``(data, target) : tuple`` option to
  :func:`load_iris` dataset
  :issue:`7049`,
  :func:`load_breast_cancer` dataset
  :issue:`7152`,
  :func:`load_digits` dataset,
  :func:`load_diabetes` dataset,
  :func:`load_linnerud` dataset,
  :func:`load_boston` dataset
  :issue:`7154` by
  :user:`Manvendra Singh<manu-chroma>`.

- Simplification of the ``clone`` function, deprecate support for estimators
  that modify parameters in ``__init__``. :issue:`5540` by `Andreas Müller`_.

- When unpickling a scikit-learn estimator in a different version than the one
  the estimator was trained with, a ``UserWarning`` is raised, see :ref:`the documentation
  on model persistence <persistence_limitations>` for more details. (:issue:`7248`)
  By `Andreas Müller`_.

Bug fixes
.........

Trees and ensembles

- Random forest, extra trees, decision trees and gradient boosting
  won't accept anymore ``min_samples_split=1`` as at least 2 samples
  are required to split a decision tree node. By `Arnaud Joly`_

- :class:`ensemble.VotingClassifier` now raises ``NotFittedError`` if ``predict``,
  ``transform`` or ``predict_proba`` are called on the non-fitted estimator.
  by `Sebastian Raschka`_.

- Fix bug where :class:`ensemble.AdaBoostClassifier` and
  :class:`ensemble.AdaBoostRegressor` would perform poorly if the
  ``random_state`` was fixed
  (:issue:`7411`). By `Joel Nothman`_.

- Fix bug in ensembles with randomization where the ensemble would not
  set ``random_state`` on base estimators in a pipeline or similar nesting.
  (:issue:`7411`). Note, results for :class:`ensemble.BaggingClassifier`
  :class:`ensemble.BaggingRegressor`, :class:`ensemble.AdaBoostClassifier`
  and :class:`ensemble.AdaBoostRegressor` will now differ from previous
  versions. By `Joel Nothman`_.

Linear, kernelized and related models

- Fixed incorrect gradient computation for ``loss='squared_epsilon_insensitive'`` in
  :class:`linear_model.SGDClassifier` and :class:`linear_model.SGDRegressor`
  (:issue:`6764`). By :user:`Wenhua Yang <geekoala>`.

- Fix bug in :class:`linear_model.LogisticRegressionCV` where
  ``solver='liblinear'`` did not accept ``class_weights='balanced``.
  (:issue:`6817`). By `Tom Dupre la Tour`_.

- Fix bug in :class:`neighbors.RadiusNeighborsClassifier` where an error
  occurred when there were outliers being labelled and a weight function
  specified (:issue:`6902`).  By
  `LeonieBorne <https://github.com/LeonieBorne>`_.

- Fix :class:`linear_model.ElasticNet` sparse decision function to match
  output with dense in the multioutput case.

Decomposition, manifold learning and clustering

- :class:`decomposition.RandomizedPCA` default number of `iterated_power` is 4 instead of 3.
  :issue:`5141` by :user:`Giorgio Patrini <giorgiop>`.

- :func:`utils.extmath.randomized_svd` performs 4 power iterations by default, instead or 0.
  In practice this is enough for obtaining a good approximation of the
  true eigenvalues/vectors in the presence of noise. When `n_components` is
  small (``< .1 * min(X.shape)``) `n_iter` is set to 7, unless the user specifies
  a higher number. This improves precision with few components.
  :issue:`5299` by :user:`Giorgio Patrini<giorgiop>`.

- Whiten/non-whiten inconsistency between components of :class:`decomposition.PCA`
  and :class:`decomposition.RandomizedPCA` (now factored into PCA, see the
  New features) is fixed. `components_` are stored with no whitening.
  :issue:`5299` by :user:`Giorgio Patrini <giorgiop>`.

- Fixed bug in :func:`manifold.spectral_embedding` where diagonal of unnormalized
  Laplacian matrix was incorrectly set to 1. :issue:`4995` by :user:`Peter Fischer <yanlend>`.

- Fixed incorrect initialization of :func:`utils.arpack.eigsh` on all
  occurrences. Affects :class:`cluster.bicluster.SpectralBiclustering`,
  :class:`decomposition.KernelPCA`, :class:`manifold.LocallyLinearEmbedding`,
  and :class:`manifold.SpectralEmbedding` (:issue:`5012`). By
  :user:`Peter Fischer <yanlend>`.

- Attribute ``explained_variance_ratio_`` calculated with the SVD solver
  of :class:`discriminant_analysis.LinearDiscriminantAnalysis` now returns
  correct results. By :user:`JPFrancoia <JPFrancoia>`

Preprocessing and feature selection

- :func:`preprocessing.data._transform_selected` now always passes a copy
  of ``X`` to transform function when ``copy=True`` (:issue:`7194`). By `Caio
  Oliveira <https://github.com/caioaao>`_.

Model evaluation and meta-estimators

- :class:`model_selection.StratifiedKFold` now raises error if all n_labels
  for individual classes is less than n_folds.
  :issue:`6182` by :user:`Devashish Deshpande <dsquareindia>`.

- Fixed bug in :class:`model_selection.StratifiedShuffleSplit`
  where train and test sample could overlap in some edge cases,
  see :issue:`6121` for
  more details. By `Loic Esteve`_.

- Fix in :class:`sklearn.model_selection.StratifiedShuffleSplit` to
  return splits of size ``train_size`` and ``test_size`` in all cases
  (:issue:`6472`). By `Andreas Müller`_.

- Cross-validation of :class:`OneVsOneClassifier` and
  :class:`OneVsRestClassifier` now works with precomputed kernels.
  :issue:`7350` by :user:`Russell Smith <rsmith54>`.

- Fix incomplete ``predict_proba`` method delegation from
  :class:`model_selection.GridSearchCV` to
  :class:`linear_model.SGDClassifier` (:issue:`7159`)
  by `Yichuan Liu <https://github.com/yl565>`_.

Metrics

- Fix bug in :func:`metrics.silhouette_score` in which clusters of
  size 1 were incorrectly scored. They should get a score of 0.
  By `Joel Nothman`_.

- Fix bug in :func:`metrics.silhouette_samples` so that it now works with
  arbitrary labels, not just those ranging from 0 to n_clusters - 1.

- Fix bug where expected and adjusted mutual information were incorrect if
  cluster contingency cells exceeded ``2**16``. By `Joel Nothman`_.

- :func:`metrics.pairwise.pairwise_distances` now converts arrays to
  boolean arrays when required in ``scipy.spatial.distance``.
  :issue:`5460` by `Tom Dupre la Tour`_.

- Fix sparse input support in :func:`metrics.silhouette_score` as well as
  example examples/text/document_clustering.py. By :user:`YenChen Lin <yenchenlin>`.

- :func:`metrics.roc_curve` and :func:`metrics.precision_recall_curve` no
  longer round ``y_score`` values when creating ROC curves; this was causing
  problems for users with very small differences in scores (:issue:`7353`).

Miscellaneous

- :func:`model_selection.tests._search._check_param_grid` now works correctly with all types
  that extends/implements `Sequence` (except string), including range (Python 3.x) and xrange
  (Python 2.x). :issue:`7323` by Viacheslav Kovalevskyi.

- :func:`utils.extmath.randomized_range_finder` is more numerically stable when many
  power iterations are requested, since it applies LU normalization by default.
  If ``n_iter<2`` numerical issues are unlikely, thus no normalization is applied.
  Other normalization options are available: ``'none', 'LU'`` and ``'QR'``.
  :issue:`5141` by :user:`Giorgio Patrini <giorgiop>`.

- Fix a bug where some formats of ``scipy.sparse`` matrix, and estimators
  with them as parameters, could not be passed to :func:`base.clone`.
  By `Loic Esteve`_.

- :func:`datasets.load_svmlight_file` now is able to read long int QID values.
  :issue:`7101` by :user:`Ibraim Ganiev <olologin>`.


API changes summary
-------------------

Linear, kernelized and related models

- ``residual_metric`` has been deprecated in :class:`linear_model.RANSACRegressor`.
  Use ``loss`` instead. By `Manoj Kumar`_.

- Access to public attributes ``.X_`` and ``.y_`` has been deprecated in
  :class:`isotonic.IsotonicRegression`. By :user:`Jonathan Arfa <jarfa>`.

Decomposition, manifold learning and clustering

- The old :class:`mixture.DPGMM` is deprecated in favor of the new
  :class:`mixture.BayesianGaussianMixture` (with the parameter
  ``weight_concentration_prior_type='dirichlet_process'``).
  The new class solves the computational
  problems of the old class and computes the Gaussian mixture with a
  Dirichlet process prior faster than before.
  :issue:`7295` by :user:`Wei Xue <xuewei4d>` and :user:`Thierry Guillemot <tguillemot>`.

- The old :class:`mixture.VBGMM` is deprecated in favor of the new
  :class:`mixture.BayesianGaussianMixture` (with the parameter
  ``weight_concentration_prior_type='dirichlet_distribution'``).
  The new class solves the computational
  problems of the old class and computes the Variational Bayesian Gaussian
  mixture faster than before.
  :issue:`6651` by :user:`Wei Xue <xuewei4d>` and :user:`Thierry Guillemot <tguillemot>`.

- The old :class:`mixture.GMM` is deprecated in favor of the new
  :class:`mixture.GaussianMixture`. The new class computes the Gaussian mixture
  faster than before and some of computational problems have been solved.
  :issue:`6666` by :user:`Wei Xue <xuewei4d>` and :user:`Thierry Guillemot <tguillemot>`.

Model evaluation and meta-estimators

- The :mod:`sklearn.cross_validation`, :mod:`sklearn.grid_search` and
  :mod:`sklearn.learning_curve` have been deprecated and the classes and
  functions have been reorganized into the :mod:`sklearn.model_selection`
  module. Ref :ref:`model_selection_changes` for more information.
  :issue:`4294` by `Raghav RV`_.

- The ``grid_scores_`` attribute of :class:`model_selection.GridSearchCV`
  and :class:`model_selection.RandomizedSearchCV` is deprecated in favor of
  the attribute ``cv_results_``.
  Ref :ref:`model_selection_changes` for more information.
  :issue:`6697` by `Raghav RV`_.

- The parameters ``n_iter`` or ``n_folds`` in old CV splitters are replaced
  by the new parameter ``n_splits`` since it can provide a consistent
  and unambiguous interface to represent the number of train-test splits.
  :issue:`7187` by :user:`YenChen Lin <yenchenlin>`.

- ``classes`` parameter was renamed to ``labels`` in
  :func:`metrics.hamming_loss`. :issue:`7260` by :user:`Sebastián Vanrell <srvanrell>`.

- The splitter classes ``LabelKFold``, ``LabelShuffleSplit``,
  ``LeaveOneLabelOut`` and ``LeavePLabelsOut`` are renamed to
  :class:`model_selection.GroupKFold`,
  :class:`model_selection.GroupShuffleSplit`,
  :class:`model_selection.LeaveOneGroupOut`
  and :class:`model_selection.LeavePGroupsOut` respectively.
  Also the parameter ``labels`` in the :func:`split` method of the newly
  renamed splitters :class:`model_selection.LeaveOneGroupOut` and
  :class:`model_selection.LeavePGroupsOut` is renamed to
  ``groups``. Additionally in :class:`model_selection.LeavePGroupsOut`,
  the parameter ``n_labels`` is renamed to ``n_groups``.
  :issue:`6660` by `Raghav RV`_.

- Error and loss names for ``scoring`` parameters are now prefixed by
  ``'neg_'``, such as ``neg_mean_squared_error``. The unprefixed versions
  are deprecated and will be removed in version 0.20.
  :issue:`7261` by :user:`Tim Head <betatim>`.

Code Contributors
-----------------
Aditya Joshi, Alejandro, Alexander Fabisch, Alexander Loginov, Alexander
Minyushkin, Alexander Rudy, Alexandre Abadie, Alexandre Abraham, Alexandre
Gramfort, Alexandre Saint, alexfields, Alvaro Ulloa, alyssaq, Amlan Kar,
Andreas Mueller, andrew giessel, Andrew Jackson, Andrew McCulloh, Andrew
Murray, Anish Shah, Arafat, Archit Sharma, Ariel Rokem, Arnaud Joly, Arnaud
Rachez, Arthur Mensch, Ash Hoover, asnt, b0noI, Behzad Tabibian, Bernardo,
Bernhard Kratzwald, Bhargav Mangipudi, blakeflei, Boyuan Deng, Brandon Carter,
Brett Naul, Brian McFee, Caio Oliveira, Camilo Lamus, Carol Willing, Cass,
CeShine Lee, Charles Truong, Chyi-Kwei Yau, CJ Carey, codevig, Colin Ni, Dan
Shiebler, Daniel, Daniel Hnyk, David Ellis, David Nicholson, David Staub, David
Thaler, David Warshaw, Davide Lasagna, Deborah, definitelyuncertain, Didi
Bar-Zev, djipey, dsquareindia, edwinENSAE, Elias Kuthe, Elvis DOHMATOB, Ethan
White, Fabian Pedregosa, Fabio Ticconi, fisache, Florian Wilhelm, Francis,
Francis O'Donovan, Gael Varoquaux, Ganiev Ibraim, ghg, Gilles Louppe, Giorgio
Patrini, Giovanni Cherubin, Giovanni Lanzani, Glenn Qian, Gordon
Mohr, govin-vatsan, Graham Clenaghan, Greg Reda, Greg Stupp, Guillaume
Lemaitre, Gustav Mörtberg, halwai, Harizo Rajaona, Harry Mavroforakis,
hashcode55, hdmetor, Henry Lin, Hobson Lane, Hugo Bowne-Anderson,
Igor Andriushchenko, Imaculate, Inki Hwang, Isaac Sijaranamual,
Ishank Gulati, Issam Laradji, Iver Jordal, jackmartin, Jacob Schreiber, Jake
Vanderplas, James Fiedler, James Routley, Jan Zikes, Janna Brettingen, jarfa, Jason
Laska, jblackburne, jeff levesque, Jeffrey Blackburne, Jeffrey04, Jeremy Hintz,
jeremynixon, Jeroen, Jessica Yung, Jill-Jênn Vie, Jimmy Jia, Jiyuan Qian, Joel
Nothman, johannah, John, John Boersma, John Kirkham, John Moeller,
jonathan.striebel, joncrall, Jordi, Joseph Munoz, Joshua Cook, JPFrancoia,
jrfiedler, JulianKahnert, juliathebrave, kaichogami, KamalakerDadi, Kenneth
Lyons, Kevin Wang, kingjr, kjell, Konstantin Podshumok, Kornel Kielczewski,
Krishna Kalyan, krishnakalyan3, Kvle Putnam, Kyle Jackson, Lars Buitinck,
ldavid, LeiG, LeightonZhang, Leland McInnes, Liang-Chi Hsieh, Lilian Besson,
lizsz, Loic Esteve, Louis Tiao, Léonie Borne, Mads Jensen, Maniteja Nandana,
Manoj Kumar, Manvendra Singh, Marco, Mario Krell, Mark Bao, Mark Szepieniec,
Martin Madsen, MartinBpr, MaryanMorel, Massil, Matheus, Mathieu Blondel,
Mathieu Dubois, Matteo, Matthias Ekman, Max Moroz, Michael Scherer, michiaki
ariga, Mikhail Korobov, Moussa Taifi, mrandrewandrade, Mridul Seth, nadya-p,
Naoya Kanai, Nate George, Nelle Varoquaux, Nelson Liu, Nick James,
NickleDave, Nico, Nicolas Goix, Nikolay Mayorov, ningchi, nlathia,
okbalefthanded, Okhlopkov, Olivier Grisel, Panos Louridas, Paul Strickland,
Perrine Letellier, pestrickland, Peter Fischer, Pieter, Ping-Yao, Chang,
practicalswift, Preston Parry, Qimu Zheng, Rachit Kansal, Raghav RV,
Ralf Gommers, Ramana.S, Rammig, Randy Olson, Rob Alexander, Robert Lutz,
Robin Schucker, Rohan Jain, Ruifeng Zheng, Ryan Yu, Rémy Léone, saihttam,
Saiwing Yeung, Sam Shleifer, Samuel St-Jean, Sartaj Singh, Sasank Chilamkurthy,
saurabh.bansod, Scott Andrews, Scott Lowe, seales, Sebastian Raschka, Sebastian
Saeger, Sebastián Vanrell, Sergei Lebedev, shagun Sodhani, shanmuga cv,
Shashank Shekhar, shawpan, shengxiduan, Shota, shuckle16, Skipper Seabold,
sklearn-ci, SmedbergM, srvanrell, Sébastien Lerique, Taranjeet, themrmax,
Thierry, Thierry Guillemot, Thomas, Thomas Hallock, Thomas Moreau, Tim Head,
tKammy, toastedcornflakes, Tom, TomDLT, Toshihiro Kamishima, tracer0tong, Trent
Hauck, trevorstephens, Tue Vo, Varun, Varun Jewalikar, Viacheslav, Vighnesh
Birodkar, Vikram, Villu Ruusmann, Vinayak Mehta, walter, waterponey, Wenhua
Yang, Wenjian Huang, Will Welch, wyseguy7, xyguo, yanlend, Yaroslav Halchenko,
yelite, Yen, YenChenLin, Yichuan Liu, Yoav Ram, Yoshiki, Zheng RuiFeng, zivori, Óscar Nájera

.. currentmodule:: sklearn

.. _changes_0_17_1:

Version 0.17.1
==============

**February 18, 2016**

Changelog
---------

Bug fixes
.........


- Upgrade vendored joblib to version 0.9.4 that fixes an important bug in
  ``joblib.Parallel`` that can silently yield to wrong results when working
  on datasets larger than 1MB:
  https://github.com/joblib/joblib/blob/0.9.4/CHANGES.rst

- Fixed reading of Bunch pickles generated with scikit-learn
  version <= 0.16. This can affect users who have already
  downloaded a dataset with scikit-learn 0.16 and are loading it
  with scikit-learn 0.17. See :issue:`6196` for
  how this affected :func:`datasets.fetch_20newsgroups`. By `Loic
  Esteve`_.

- Fixed a bug that prevented using ROC AUC score to perform grid search on
  several CPU / cores on large arrays. See :issue:`6147`
  By `Olivier Grisel`_.

- Fixed a bug that prevented to properly set the ``presort`` parameter
  in :class:`ensemble.GradientBoostingRegressor`. See :issue:`5857`
  By Andrew McCulloh.

- Fixed a joblib error when evaluating the perplexity of a
  :class:`decomposition.LatentDirichletAllocation` model. See :issue:`6258`
  By Chyi-Kwei Yau.


.. _changes_0_17:

Version 0.17
============

**November 5, 2015**

Changelog
---------

New features
............

- All the Scaler classes but :class:`preprocessing.RobustScaler` can be fitted online by
  calling `partial_fit`. By :user:`Giorgio Patrini <giorgiop>`.

- The new class :class:`ensemble.VotingClassifier` implements a
  "majority rule" / "soft voting" ensemble classifier to combine
  estimators for classification. By `Sebastian Raschka`_.

- The new class :class:`preprocessing.RobustScaler` provides an
  alternative to :class:`preprocessing.StandardScaler` for feature-wise
  centering and range normalization that is robust to outliers.
  By :user:`Thomas Unterthiner <untom>`.

- The new class :class:`preprocessing.MaxAbsScaler` provides an
  alternative to :class:`preprocessing.MinMaxScaler` for feature-wise
  range normalization when the data is already centered or sparse.
  By :user:`Thomas Unterthiner <untom>`.

- The new class :class:`preprocessing.FunctionTransformer` turns a Python
  function into a ``Pipeline``-compatible transformer object.
  By Joe Jevnik.

- The new classes :class:`cross_validation.LabelKFold` and
  :class:`cross_validation.LabelShuffleSplit` generate train-test folds,
  respectively similar to :class:`cross_validation.KFold` and
  :class:`cross_validation.ShuffleSplit`, except that the folds are
  conditioned on a label array. By `Brian McFee`_, :user:`Jean
  Kossaifi <JeanKossaifi>` and `Gilles Louppe`_.

- :class:`decomposition.LatentDirichletAllocation` implements the Latent
  Dirichlet Allocation topic model with online  variational
  inference. By :user:`Chyi-Kwei Yau <chyikwei>`, with code based on an implementation
  by Matt Hoffman. (:issue:`3659`)

- The new solver ``sag`` implements a Stochastic Average Gradient descent
  and is available in both :class:`linear_model.LogisticRegression` and
  :class:`linear_model.Ridge`. This solver is very efficient for large
  datasets. By :user:`Danny Sullivan <dsullivan7>` and `Tom Dupre la Tour`_.
  (:issue:`4738`)

- The new solver ``cd`` implements a Coordinate Descent in
  :class:`decomposition.NMF`. Previous solver based on Projected Gradient is
  still available setting new parameter ``solver`` to ``pg``, but is
  deprecated and will be removed in 0.19, along with
  :class:`decomposition.ProjectedGradientNMF` and parameters ``sparseness``,
  ``eta``, ``beta`` and ``nls_max_iter``. New parameters ``alpha`` and
  ``l1_ratio`` control L1 and L2 regularization, and ``shuffle`` adds a
  shuffling step in the ``cd`` solver.
  By `Tom Dupre la Tour`_ and `Mathieu Blondel`_.

Enhancements
............
- :class:`manifold.TSNE` now supports approximate optimization via the
  Barnes-Hut method, leading to much faster fitting. By Christopher Erick Moody.
  (:issue:`4025`)

- :class:`cluster.mean_shift_.MeanShift` now supports parallel execution,
  as implemented in the ``mean_shift`` function. By :user:`Martino
  Sorbaro <martinosorb>`.

- :class:`naive_bayes.GaussianNB` now supports fitting with ``sample_weight``.
  By `Jan Hendrik Metzen`_.

- :class:`dummy.DummyClassifier` now supports a prior fitting strategy.
  By `Arnaud Joly`_.

- Added a ``fit_predict`` method for :class:`mixture.GMM` and subclasses.
  By :user:`Cory Lorenz <clorenz7>`.

- Added the :func:`metrics.label_ranking_loss` metric.
  By `Arnaud Joly`_.

- Added the :func:`metrics.cohen_kappa_score` metric.

- Added a ``warm_start`` constructor parameter to the bagging ensemble
  models to increase the size of the ensemble. By :user:`Tim Head <betatim>`.

- Added option to use multi-output regression metrics without averaging.
  By Konstantin Shmelkov and :user:`Michael Eickenberg<eickenberg>`.

- Added ``stratify`` option to :func:`cross_validation.train_test_split`
  for stratified splitting. By Miroslav Batchkarov.

- The :func:`tree.export_graphviz` function now supports aesthetic
  improvements for :class:`tree.DecisionTreeClassifier` and
  :class:`tree.DecisionTreeRegressor`, including options for coloring nodes
  by their majority class or impurity, showing variable names, and using
  node proportions instead of raw sample counts. By `Trevor Stephens`_.

- Improved speed of ``newton-cg`` solver in
  :class:`linear_model.LogisticRegression`, by avoiding loss computation.
  By `Mathieu Blondel`_ and `Tom Dupre la Tour`_.

- The ``class_weight="auto"`` heuristic in classifiers supporting
  ``class_weight`` was deprecated and replaced by the ``class_weight="balanced"``
  option, which has a simpler formula and interpretation.
  By `Hanna Wallach`_ and `Andreas Müller`_.

- Add ``class_weight`` parameter to automatically weight samples by class
  frequency for :class:`linear_model.PassiveAgressiveClassifier`. By
  `Trevor Stephens`_.

- Added backlinks from the API reference pages to the user guide. By
  `Andreas Müller`_.

- The ``labels`` parameter to :func:`sklearn.metrics.f1_score`,
  :func:`sklearn.metrics.fbeta_score`,
  :func:`sklearn.metrics.recall_score` and
  :func:`sklearn.metrics.precision_score` has been extended.
  It is now possible to ignore one or more labels, such as where
  a multiclass problem has a majority class to ignore. By `Joel Nothman`_.

- Add ``sample_weight`` support to :class:`linear_model.RidgeClassifier`.
  By `Trevor Stephens`_.

- Provide an option for sparse output from
  :func:`sklearn.metrics.pairwise.cosine_similarity`. By
  :user:`Jaidev Deshpande <jaidevd>`.

- Add :func:`minmax_scale` to provide a function interface for
  :class:`MinMaxScaler`. By :user:`Thomas Unterthiner <untom>`.

- ``dump_svmlight_file`` now handles multi-label datasets.
  By Chih-Wei Chang.

- RCV1 dataset loader (:func:`sklearn.datasets.fetch_rcv1`).
  By `Tom Dupre la Tour`_.

- The "Wisconsin Breast Cancer" classical two-class classification dataset
  is now included in scikit-learn, available with
  :func:`sklearn.dataset.load_breast_cancer`.

- Upgraded to joblib 0.9.3 to benefit from the new automatic batching of
  short tasks. This makes it possible for scikit-learn to benefit from
  parallelism when many very short tasks are executed in parallel, for
  instance by the :class:`grid_search.GridSearchCV` meta-estimator
  with ``n_jobs > 1`` used with a large grid of parameters on a small
  dataset. By `Vlad Niculae`_, `Olivier Grisel`_ and `Loic Esteve`_.

- For more details about changes in joblib 0.9.3 see the release notes:
  https://github.com/joblib/joblib/blob/master/CHANGES.rst#release-093

- Improved speed (3 times per iteration) of
  :class:`decomposition.DictLearning` with coordinate descent method
  from :class:`linear_model.Lasso`. By :user:`Arthur Mensch <arthurmensch>`.

- Parallel processing (threaded) for queries of nearest neighbors
  (using the ball-tree) by Nikolay Mayorov.

- Allow :func:`datasets.make_multilabel_classification` to output
  a sparse ``y``. By Kashif Rasul.

- :class:`cluster.DBSCAN` now accepts a sparse matrix of precomputed
  distances, allowing memory-efficient distance precomputation. By
  `Joel Nothman`_.

- :class:`tree.DecisionTreeClassifier` now exposes an ``apply`` method
  for retrieving the leaf indices samples are predicted as. By
  :user:`Daniel Galvez <galv>` and `Gilles Louppe`_.

- Speed up decision tree regressors, random forest regressors, extra trees
  regressors and gradient boosting estimators by computing a proxy
  of the impurity improvement during the tree growth. The proxy quantity is
  such that the split that maximizes this value also maximizes the impurity
  improvement. By `Arnaud Joly`_, :user:`Jacob Schreiber <jmschrei>`
  and `Gilles Louppe`_.

- Speed up tree based methods by reducing the number of computations needed
  when computing the impurity measure taking into account linear
  relationship of the computed statistics. The effect is particularly
  visible with extra trees and on datasets with categorical or sparse
  features. By `Arnaud Joly`_.

- :class:`ensemble.GradientBoostingRegressor` and
  :class:`ensemble.GradientBoostingClassifier` now expose an ``apply``
  method for retrieving the leaf indices each sample ends up in under
  each try. By :user:`Jacob Schreiber <jmschrei>`.

- Add ``sample_weight`` support to :class:`linear_model.LinearRegression`.
  By Sonny Hu. (:issue:`#4881`)

- Add ``n_iter_without_progress`` to :class:`manifold.TSNE` to control
  the stopping criterion. By Santi Villalba. (:issue:`5186`)

- Added optional parameter ``random_state`` in :class:`linear_model.Ridge`
  , to set the seed of the pseudo random generator used in ``sag`` solver. By `Tom Dupre la Tour`_.

- Added optional parameter ``warm_start`` in
  :class:`linear_model.LogisticRegression`. If set to True, the solvers
  ``lbfgs``, ``newton-cg`` and ``sag`` will be initialized with the
  coefficients computed in the previous fit. By `Tom Dupre la Tour`_.

- Added ``sample_weight`` support to :class:`linear_model.LogisticRegression` for
  the ``lbfgs``, ``newton-cg``, and ``sag`` solvers. By `Valentin Stolbunov`_.
  Support added to the ``liblinear`` solver. By `Manoj Kumar`_.

- Added optional parameter ``presort`` to :class:`ensemble.GradientBoostingRegressor`
  and :class:`ensemble.GradientBoostingClassifier`, keeping default behavior
  the same. This allows gradient boosters to turn off presorting when building
  deep trees or using sparse data. By :user:`Jacob Schreiber <jmschrei>`.

- Altered :func:`metrics.roc_curve` to drop unnecessary thresholds by
  default. By :user:`Graham Clenaghan <gclenaghan>`.

- Added :class:`feature_selection.SelectFromModel` meta-transformer which can
  be used along with estimators that have `coef_` or `feature_importances_`
  attribute to select important features of the input data. By
  :user:`Maheshakya Wijewardena <maheshakya>`, `Joel Nothman`_ and `Manoj Kumar`_.

- Added :func:`metrics.pairwise.laplacian_kernel`.  By `Clyde Fare <https://github.com/Clyde-fare>`_.

- :class:`covariance.GraphLasso` allows separate control of the convergence criterion
  for the Elastic-Net subproblem via  the ``enet_tol`` parameter.

- Improved verbosity in :class:`decomposition.DictionaryLearning`.

- :class:`ensemble.RandomForestClassifier` and
  :class:`ensemble.RandomForestRegressor` no longer explicitly store the
  samples used in bagging, resulting in a much reduced memory footprint for
  storing random forest models.

- Added ``positive`` option to :class:`linear_model.Lars` and
  :func:`linear_model.lars_path` to force coefficients to be positive.
  (:issue:`5131`)

- Added the ``X_norm_squared`` parameter to :func:`metrics.pairwise.euclidean_distances`
  to provide precomputed squared norms for ``X``.

- Added the ``fit_predict`` method to :class:`pipeline.Pipeline`.

- Added the :func:`preprocessing.min_max_scale` function.

Bug fixes
.........

- Fixed non-determinism in :class:`dummy.DummyClassifier` with sparse
  multi-label output. By `Andreas Müller`_.

- Fixed the output shape of :class:`linear_model.RANSACRegressor` to
  ``(n_samples, )``. By `Andreas Müller`_.

- Fixed bug in :class:`decomposition.DictLearning` when ``n_jobs < 0``. By
  `Andreas Müller`_.

- Fixed bug where :class:`grid_search.RandomizedSearchCV` could consume a
  lot of memory for large discrete grids. By `Joel Nothman`_.

- Fixed bug in :class:`linear_model.LogisticRegressionCV` where `penalty` was ignored
  in the final fit. By `Manoj Kumar`_.

- Fixed bug in :class:`ensemble.forest.ForestClassifier` while computing
  oob_score and X is a sparse.csc_matrix. By :user:`Ankur Ankan <ankurankan>`.

- All regressors now consistently handle and warn when given ``y`` that is of
  shape ``(n_samples, 1)``. By `Andreas Müller`_ and Henry Lin.
  (:issue:`5431`)

- Fix in :class:`cluster.KMeans` cluster reassignment for sparse input by
  `Lars Buitinck`_.

- Fixed a bug in :class:`lda.LDA` that could cause asymmetric covariance
  matrices when using shrinkage. By `Martin Billinger`_.

- Fixed :func:`cross_validation.cross_val_predict` for estimators with
  sparse predictions. By Buddha Prakash.

- Fixed the ``predict_proba`` method of :class:`linear_model.LogisticRegression`
  to use soft-max instead of one-vs-rest normalization. By `Manoj Kumar`_.
  (:issue:`5182`)

- Fixed the :func:`partial_fit` method of :class:`linear_model.SGDClassifier`
  when called with ``average=True``. By :user:`Andrew Lamb <andylamb>`.
  (:issue:`5282`)

- Dataset fetchers use different filenames under Python 2 and Python 3 to
  avoid pickling compatibility issues. By `Olivier Grisel`_.
  (:issue:`5355`)

- Fixed a bug in :class:`naive_bayes.GaussianNB` which caused classification
  results to depend on scale. By `Jake Vanderplas`_.

- Fixed temporarily :class:`linear_model.Ridge`, which was incorrect
  when fitting the intercept in the case of sparse data. The fix
  automatically changes the solver to 'sag' in this case.
  :issue:`5360` by `Tom Dupre la Tour`_.

- Fixed a performance bug in :class:`decomposition.RandomizedPCA` on data
  with a large number of features and fewer samples. (:issue:`4478`)
  By `Andreas Müller`_, `Loic Esteve`_ and :user:`Giorgio Patrini <giorgiop>`.

- Fixed bug in :class:`cross_decomposition.PLS` that yielded unstable and
  platform dependent output, and failed on `fit_transform`.
  By :user:`Arthur Mensch <arthurmensch>`.

- Fixes to the ``Bunch`` class used to store datasets.

- Fixed :func:`ensemble.plot_partial_dependence` ignoring the
  ``percentiles`` parameter.

- Providing a ``set`` as vocabulary in ``CountVectorizer`` no longer
  leads to inconsistent results when pickling.

- Fixed the conditions on when a precomputed Gram matrix needs to
  be recomputed in :class:`linear_model.LinearRegression`,
  :class:`linear_model.OrthogonalMatchingPursuit`,
  :class:`linear_model.Lasso` and :class:`linear_model.ElasticNet`.

- Fixed inconsistent memory layout in the coordinate descent solver
  that affected :class:`linear_model.DictionaryLearning` and
  :class:`covariance.GraphLasso`. (:issue:`5337`)
  By `Olivier Grisel`_.

- :class:`manifold.LocallyLinearEmbedding` no longer ignores the ``reg``
  parameter.

- Nearest Neighbor estimators with custom distance metrics can now be pickled.
  (:issue:`4362`)

- Fixed a bug in :class:`pipeline.FeatureUnion` where ``transformer_weights``
  were not properly handled when performing grid-searches.

- Fixed a bug in :class:`linear_model.LogisticRegression` and
  :class:`linear_model.LogisticRegressionCV` when using
  ``class_weight='balanced'```or ``class_weight='auto'``.
  By `Tom Dupre la Tour`_.

- Fixed bug :issue:`5495` when
  doing OVR(SVC(decision_function_shape="ovr")). Fixed by
  :user:`Elvis Dohmatob <dohmatob>`.


API changes summary
-------------------
- Attribute `data_min`, `data_max` and `data_range` in
  :class:`preprocessing.MinMaxScaler` are deprecated and won't be available
  from 0.19. Instead, the class now exposes `data_min_`, `data_max_`
  and `data_range_`. By :user:`Giorgio Patrini <giorgiop>`.

- All Scaler classes now have an `scale_` attribute, the feature-wise
  rescaling applied by their `transform` methods. The old attribute `std_`
  in :class:`preprocessing.StandardScaler` is deprecated and superseded
  by `scale_`; it won't be available in 0.19. By :user:`Giorgio Patrini <giorgiop>`.

- :class:`svm.SVC`` and :class:`svm.NuSVC` now have an ``decision_function_shape``
  parameter to make their decision function of shape ``(n_samples, n_classes)``
  by setting ``decision_function_shape='ovr'``. This will be the default behavior
  starting in 0.19. By `Andreas Müller`_.

- Passing 1D data arrays as input to estimators is now deprecated as it
  caused confusion in how the array elements should be interpreted
  as features or as samples. All data arrays are now expected
  to be explicitly shaped ``(n_samples, n_features)``.
  By :user:`Vighnesh Birodkar <vighneshbirodkar>`.

- :class:`lda.LDA` and :class:`qda.QDA` have been moved to
  :class:`discriminant_analysis.LinearDiscriminantAnalysis` and
  :class:`discriminant_analysis.QuadraticDiscriminantAnalysis`.

- The ``store_covariance`` and ``tol`` parameters have been moved from
  the fit method to the constructor in
  :class:`discriminant_analysis.LinearDiscriminantAnalysis` and the
  ``store_covariances`` and ``tol`` parameters have been moved from the
  fit method to the constructor in
  :class:`discriminant_analysis.QuadraticDiscriminantAnalysis`.

- Models inheriting from ``_LearntSelectorMixin`` will no longer support the
  transform methods. (i.e,  RandomForests, GradientBoosting, LogisticRegression,
  DecisionTrees, SVMs and SGD related models). Wrap these models around the
  metatransfomer :class:`feature_selection.SelectFromModel` to remove
  features (according to `coefs_` or `feature_importances_`)
  which are below a certain threshold value instead.

- :class:`cluster.KMeans` re-runs cluster-assignments in case of non-convergence,
  to ensure consistency of ``predict(X)`` and ``labels_``. By
  :user:`Vighnesh Birodkar <vighneshbirodkar>`.

- Classifier and Regressor models are now tagged as such using the
  ``_estimator_type`` attribute.

- Cross-validation iterators always provide indices into training and test set,
  not boolean masks.

- The ``decision_function`` on all regressors was deprecated and will be
  removed in 0.19.  Use ``predict`` instead.

- :func:`datasets.load_lfw_pairs` is deprecated and will be removed in 0.19.
  Use :func:`datasets.fetch_lfw_pairs` instead.

- The deprecated ``hmm`` module was removed.

- The deprecated ``Bootstrap`` cross-validation iterator was removed.

- The deprecated ``Ward`` and ``WardAgglomerative`` classes have been removed.
  Use :class:`clustering.AgglomerativeClustering` instead.

- :func:`cross_validation.check_cv` is now a public function.

- The property ``residues_`` of :class:`linear_model.LinearRegression` is deprecated
  and will be removed in 0.19.

- The deprecated ``n_jobs`` parameter of :class:`linear_model.LinearRegression` has been moved
  to the constructor.

- Removed deprecated ``class_weight`` parameter from :class:`linear_model.SGDClassifier`'s ``fit``
  method. Use the construction parameter instead.

- The deprecated support for the sequence of sequences (or list of lists) multilabel
  format was removed. To convert to and from the supported binary
  indicator matrix format, use
  :class:`MultiLabelBinarizer <preprocessing.MultiLabelBinarizer>`.

- The behavior of calling the ``inverse_transform`` method of ``Pipeline.pipeline`` will
  change in 0.19. It will no longer reshape one-dimensional input to two-dimensional input.

- The deprecated attributes ``indicator_matrix_``, ``multilabel_`` and ``classes_`` of
  :class:`preprocessing.LabelBinarizer` were removed.

- Using ``gamma=0`` in :class:`svm.SVC` and :class:`svm.SVR` to automatically set the
  gamma to ``1. / n_features`` is deprecated and will be removed in 0.19.
  Use ``gamma="auto"`` instead.

Code Contributors
-----------------
Aaron Schumacher, Adithya Ganesh, akitty, Alexandre Gramfort, Alexey Grigorev,
Ali Baharev, Allen Riddell, Ando Saabas, Andreas Mueller, Andrew Lamb, Anish
Shah, Ankur Ankan, Anthony Erlinger, Ari Rouvinen, Arnaud Joly, Arnaud Rachez,
Arthur Mensch, banilo, Barmaley.exe, benjaminirving, Boyuan Deng, Brett Naul,
Brian McFee, Buddha Prakash, Chi Zhang, Chih-Wei Chang, Christof Angermueller,
Christoph Gohlke, Christophe Bourguignat, Christopher Erick Moody, Chyi-Kwei
Yau, Cindy Sridharan, CJ Carey, Clyde-fare, Cory Lorenz, Dan Blanchard, Daniel
Galvez, Daniel Kronovet, Danny Sullivan, Data1010, David, David D Lowe, David
Dotson, djipey, Dmitry Spikhalskiy, Donne Martin, Dougal J. Sutherland, Dougal
Sutherland, edson duarte, Eduardo Caro, Eric Larson, Eric Martin, Erich
Schubert, Fernando Carrillo, Frank C. Eckert, Frank Zalkow, Gael Varoquaux,
Ganiev Ibraim, Gilles Louppe, Giorgio Patrini, giorgiop, Graham Clenaghan,
Gryllos Prokopis, gwulfs, Henry Lin, Hsuan-Tien Lin, Immanuel Bayer, Ishank
Gulati, Jack Martin, Jacob Schreiber, Jaidev Deshpande, Jake Vanderplas, Jan
Hendrik Metzen, Jean Kossaifi, Jeffrey04, Jeremy, jfraj, Jiali Mei,
Joe Jevnik, Joel Nothman, John Kirkham, John Wittenauer, Joseph, Joshua Loyal,
Jungkook Park, KamalakerDadi, Kashif Rasul, Keith Goodman, Kian Ho, Konstantin
Shmelkov, Kyler Brown, Lars Buitinck, Lilian Besson, Loic Esteve, Louis Tiao,
maheshakya, Maheshakya Wijewardena, Manoj Kumar, MarkTab marktab.net, Martin
Ku, Martin Spacek, MartinBpr, martinosorb, MaryanMorel, Masafumi Oyamada,
Mathieu Blondel, Matt Krump, Matti Lyra, Maxim Kolganov, mbillinger, mhg,
Michael Heilman, Michael Patterson, Miroslav Batchkarov, Nelle Varoquaux,
Nicolas, Nikolay Mayorov, Olivier Grisel, Omer Katz, Óscar Nájera, Pauli
Virtanen, Peter Fischer, Peter Prettenhofer, Phil Roth, pianomania, Preston
Parry, Raghav RV, Rob Zinkov, Robert Layton, Rohan Ramanath, Saket Choudhary,
Sam Zhang, santi, saurabh.bansod, scls19fr, Sebastian Raschka, Sebastian
Saeger, Shivan Sornarajah, SimonPL, sinhrks, Skipper Seabold, Sonny Hu, sseg,
Stephen Hoover, Steven De Gryze, Steven Seguin, Theodore Vasiloudis, Thomas
Unterthiner, Tiago Freitas Pereira, Tian Wang, Tim Head, Timothy Hopper,
tokoroten, Tom Dupré la Tour, Trevor Stephens, Valentin Stolbunov, Vighnesh
Birodkar, Vinayak Mehta, Vincent, Vincent Michel, vstolbunov, wangz10, Wei Xue,
Yucheng Low, Yury Zhauniarovich, Zac Stewart, zhai_pro, Zichen Wang

.. _changes_0_1_16:

Version 0.16.1
===============

**April 14, 2015**

Changelog
---------

Bug fixes
.........

- Allow input data larger than ``block_size`` in
  :class:`covariance.LedoitWolf` by `Andreas Müller`_.

- Fix a bug in :class:`isotonic.IsotonicRegression` deduplication that
  caused unstable result in :class:`calibration.CalibratedClassifierCV` by
  `Jan Hendrik Metzen`_.

- Fix sorting of labels in func:`preprocessing.label_binarize` by Michael Heilman.

- Fix several stability and convergence issues in
  :class:`cross_decomposition.CCA` and
  :class:`cross_decomposition.PLSCanonical` by `Andreas Müller`_

- Fix a bug in :class:`cluster.KMeans` when ``precompute_distances=False``
  on fortran-ordered data.

- Fix a speed regression in :class:`ensemble.RandomForestClassifier`'s ``predict``
  and ``predict_proba`` by `Andreas Müller`_.

- Fix a regression where ``utils.shuffle`` converted lists and dataframes to arrays, by `Olivier Grisel`_

.. _changes_0_16:

Version 0.16
============

**March 26, 2015**

Highlights
-----------

- Speed improvements (notably in :class:`cluster.DBSCAN`), reduced memory
  requirements, bug-fixes and better default settings.

- Multinomial Logistic regression and a path algorithm in
  :class:`linear_model.LogisticRegressionCV`.

- Out-of core learning of PCA via :class:`decomposition.IncrementalPCA`.

- Probability callibration of classifiers using
  :class:`calibration.CalibratedClassifierCV`.

- :class:`cluster.Birch` clustering method for large-scale datasets.

- Scalable approximate nearest neighbors search with Locality-sensitive
  hashing forests in :class:`neighbors.LSHForest`.

- Improved error messages and better validation when using malformed input data.

- More robust integration with pandas dataframes.

Changelog
---------

New features
............

- The new :class:`neighbors.LSHForest` implements locality-sensitive hashing
  for approximate nearest neighbors search. By :user:`Maheshakya Wijewardena<maheshakya>`.

- Added :class:`svm.LinearSVR`. This class uses the liblinear implementation
  of Support Vector Regression which is much faster for large
  sample sizes than :class:`svm.SVR` with linear kernel. By
  `Fabian Pedregosa`_ and Qiang Luo.

- Incremental fit for :class:`GaussianNB <naive_bayes.GaussianNB>`.

- Added ``sample_weight`` support to :class:`dummy.DummyClassifier` and
  :class:`dummy.DummyRegressor`. By `Arnaud Joly`_.

- Added the :func:`metrics.label_ranking_average_precision_score` metrics.
  By `Arnaud Joly`_.

- Add the :func:`metrics.coverage_error` metrics. By `Arnaud Joly`_.

- Added :class:`linear_model.LogisticRegressionCV`. By
  `Manoj Kumar`_, `Fabian Pedregosa`_, `Gael Varoquaux`_
  and `Alexandre Gramfort`_.

- Added ``warm_start`` constructor parameter to make it possible for any
  trained forest model to grow additional trees incrementally. By
  :user:`Laurent Direr<ldirer>`.

- Added ``sample_weight`` support to :class:`ensemble.GradientBoostingClassifier` and
  :class:`ensemble.GradientBoostingRegressor`. By `Peter Prettenhofer`_.

- Added :class:`decomposition.IncrementalPCA`, an implementation of the PCA
  algorithm that supports out-of-core learning with a ``partial_fit``
  method. By `Kyle Kastner`_.

- Averaged SGD for :class:`SGDClassifier <linear_model.SGDClassifier>`
  and :class:`SGDRegressor <linear_model.SGDRegressor>` By
  :user:`Danny Sullivan <dsullivan7>`.

- Added :func:`cross_val_predict <cross_validation.cross_val_predict>`
  function which computes cross-validated estimates. By `Luis Pedro Coelho`_

- Added :class:`linear_model.TheilSenRegressor`, a robust
  generalized-median-based estimator. By :user:`Florian Wilhelm <FlorianWilhelm>`.

- Added :func:`metrics.median_absolute_error`, a robust metric.
  By `Gael Varoquaux`_ and :user:`Florian Wilhelm <FlorianWilhelm>`.

- Add :class:`cluster.Birch`, an online clustering algorithm. By
  `Manoj Kumar`_, `Alexandre Gramfort`_ and `Joel Nothman`_.

- Added shrinkage support to :class:`discriminant_analysis.LinearDiscriminantAnalysis`
  using two new solvers. By :user:`Clemens Brunner <cle1109>` and `Martin Billinger`_.

- Added :class:`kernel_ridge.KernelRidge`, an implementation of
  kernelized ridge regression.
  By `Mathieu Blondel`_ and `Jan Hendrik Metzen`_.

- All solvers in :class:`linear_model.Ridge` now support `sample_weight`.
  By `Mathieu Blondel`_.

- Added :class:`cross_validation.PredefinedSplit` cross-validation
  for fixed user-provided cross-validation folds.
  By :user:`Thomas Unterthiner <untom>`.

- Added :class:`calibration.CalibratedClassifierCV`, an approach for
  calibrating the predicted probabilities of a classifier.
  By `Alexandre Gramfort`_, `Jan Hendrik Metzen`_, `Mathieu Blondel`_
  and :user:`Balazs Kegl <kegl>`.


Enhancements
............

- Add option ``return_distance`` in :func:`hierarchical.ward_tree`
  to return distances between nodes for both structured and unstructured
  versions of the algorithm. By `Matteo Visconti di Oleggio Castello`_.
  The same option was added in :func:`hierarchical.linkage_tree`.
  By `Manoj Kumar`_

- Add support for sample weights in scorer objects.  Metrics with sample
  weight support will automatically benefit from it. By `Noel Dawe`_ and
  `Vlad Niculae`_.

- Added ``newton-cg`` and `lbfgs` solver support in
  :class:`linear_model.LogisticRegression`. By `Manoj Kumar`_.

- Add ``selection="random"`` parameter to implement stochastic coordinate
  descent for :class:`linear_model.Lasso`, :class:`linear_model.ElasticNet`
  and related. By `Manoj Kumar`_.

- Add ``sample_weight`` parameter to
  :func:`metrics.jaccard_similarity_score` and :func:`metrics.log_loss`.
  By :user:`Jatin Shah <jatinshah>`.

- Support sparse multilabel indicator representation in
  :class:`preprocessing.LabelBinarizer` and
  :class:`multiclass.OneVsRestClassifier` (by :user:`Hamzeh Alsalhi <hamsal>` with thanks
  to Rohit Sivaprasad), as well as evaluation metrics (by
  `Joel Nothman`_).

- Add ``sample_weight`` parameter to `metrics.jaccard_similarity_score`.
  By `Jatin Shah`.

- Add support for multiclass in `metrics.hinge_loss`. Added ``labels=None``
  as optional parameter. By `Saurabh Jha`.

- Add ``sample_weight`` parameter to `metrics.hinge_loss`.
  By `Saurabh Jha`.

- Add ``multi_class="multinomial"`` option in
  :class:`linear_model.LogisticRegression` to implement a Logistic
  Regression solver that minimizes the cross-entropy or multinomial loss
  instead of the default One-vs-Rest setting. Supports `lbfgs` and
  `newton-cg` solvers. By `Lars Buitinck`_ and `Manoj Kumar`_. Solver option
  `newton-cg` by Simon Wu.

- ``DictVectorizer`` can now perform ``fit_transform`` on an iterable in a
  single pass, when giving the option ``sort=False``. By :user:`Dan
  Blanchard <dan-blanchard>`.

- :class:`GridSearchCV` and :class:`RandomizedSearchCV` can now be
  configured to work with estimators that may fail and raise errors on
  individual folds. This option is controlled by the `error_score`
  parameter. This does not affect errors raised on re-fit. By
  :user:`Michal Romaniuk <romaniukm>`.

- Add ``digits`` parameter to `metrics.classification_report` to allow
  report to show different precision of floating point numbers. By
  :user:`Ian Gilmore <agileminor>`.

- Add a quantile prediction strategy to the :class:`dummy.DummyRegressor`.
  By :user:`Aaron Staple <staple>`.

- Add ``handle_unknown`` option to :class:`preprocessing.OneHotEncoder` to
  handle unknown categorical features more gracefully during transform.
  By `Manoj Kumar`_.

- Added support for sparse input data to decision trees and their ensembles.
  By `Fares Hedyati`_ and `Arnaud Joly`_.

- Optimized :class:`cluster.AffinityPropagation` by reducing the number of
  memory allocations of large temporary data-structures. By `Antony Lee`_.

- Parellization of the computation of feature importances in random forest.
  By `Olivier Grisel`_ and `Arnaud Joly`_.

- Add ``n_iter_`` attribute to estimators that accept a ``max_iter`` attribute
  in their constructor. By `Manoj Kumar`_.

- Added decision function for :class:`multiclass.OneVsOneClassifier`
  By `Raghav RV`_ and :user:`Kyle Beauchamp <kyleabeauchamp>`.

- :func:`neighbors.kneighbors_graph` and :func:`radius_neighbors_graph`
  support non-Euclidean metrics. By `Manoj Kumar`_

- Parameter ``connectivity`` in :class:`cluster.AgglomerativeClustering`
  and family now accept callables that return a connectivity matrix.
  By `Manoj Kumar`_.

- Sparse support for :func:`paired_distances`. By `Joel Nothman`_.

- :class:`cluster.DBSCAN` now supports sparse input and sample weights and
  has been optimized: the inner loop has been rewritten in Cython and
  radius neighbors queries are now computed in batch. By `Joel Nothman`_
  and `Lars Buitinck`_.

- Add ``class_weight`` parameter to automatically weight samples by class
  frequency for :class:`ensemble.RandomForestClassifier`,
  :class:`tree.DecisionTreeClassifier`, :class:`ensemble.ExtraTreesClassifier`
  and :class:`tree.ExtraTreeClassifier`. By `Trevor Stephens`_.

- :class:`grid_search.RandomizedSearchCV` now does sampling without
  replacement if all parameters are given as lists. By `Andreas Müller`_.

- Parallelized calculation of :func:`pairwise_distances` is now supported
  for scipy metrics and custom callables. By `Joel Nothman`_.

- Allow the fitting and scoring of all clustering algorithms in
  :class:`pipeline.Pipeline`. By `Andreas Müller`_.

- More robust seeding and improved error messages in :class:`cluster.MeanShift`
  by `Andreas Müller`_.

- Make the stopping criterion for :class:`mixture.GMM`,
  :class:`mixture.DPGMM` and :class:`mixture.VBGMM` less dependent on the
  number of samples by thresholding the average log-likelihood change
  instead of its sum over all samples. By `Hervé Bredin`_.

- The outcome of :func:`manifold.spectral_embedding` was made deterministic
  by flipping the sign of eigenvectors. By :user:`Hasil Sharma <Hasil-Sharma>`.

- Significant performance and memory usage improvements in
  :class:`preprocessing.PolynomialFeatures`. By `Eric Martin`_.

- Numerical stability improvements for :class:`preprocessing.StandardScaler`
  and :func:`preprocessing.scale`. By `Nicolas Goix`_

- :class:`svm.SVC` fitted on sparse input now implements ``decision_function``.
  By `Rob Zinkov`_ and `Andreas Müller`_.

- :func:`cross_validation.train_test_split` now preserves the input type,
  instead of converting to numpy arrays.


Documentation improvements
..........................

- Added example of using :class:`FeatureUnion` for heterogeneous input.
  By :user:`Matt Terry <mrterry>`

- Documentation on scorers was improved, to highlight the handling of loss
  functions. By :user:`Matt Pico <MattpSoftware>`.

- A discrepancy between liblinear output and scikit-learn's wrappers
  is now noted. By `Manoj Kumar`_.

- Improved documentation generation: examples referring to a class or
  function are now shown in a gallery on the class/function's API reference
  page. By `Joel Nothman`_.

- More explicit documentation of sample generators and of data
  transformation. By `Joel Nothman`_.

- :class:`sklearn.neighbors.BallTree` and :class:`sklearn.neighbors.KDTree`
  used to point to empty pages stating that they are aliases of BinaryTree.
  This has been fixed to show the correct class docs. By `Manoj Kumar`_.

- Added silhouette plots for analysis of KMeans clustering using
  :func:`metrics.silhouette_samples` and :func:`metrics.silhouette_score`.
  See :ref:`sphx_glr_auto_examples_cluster_plot_kmeans_silhouette_analysis.py`

Bug fixes
.........
- Metaestimators now support ducktyping for the presence of ``decision_function``,
  ``predict_proba`` and other methods. This fixes behavior of
  :class:`grid_search.GridSearchCV`,
  :class:`grid_search.RandomizedSearchCV`, :class:`pipeline.Pipeline`,
  :class:`feature_selection.RFE`, :class:`feature_selection.RFECV` when nested.
  By `Joel Nothman`_

- The ``scoring`` attribute of grid-search and cross-validation methods is no longer
  ignored when a :class:`grid_search.GridSearchCV` is given as a base estimator or
  the base estimator doesn't have predict.

- The function :func:`hierarchical.ward_tree` now returns the children in
  the same order for both the structured and unstructured versions. By
  `Matteo Visconti di Oleggio Castello`_.

- :class:`feature_selection.RFECV` now correctly handles cases when
  ``step`` is not equal to 1. By :user:`Nikolay Mayorov <nmayorov>`

- The :class:`decomposition.PCA` now undoes whitening in its
  ``inverse_transform``. Also, its ``components_`` now always have unit
  length. By :user:`Michael Eickenberg <eickenberg>`.

- Fix incomplete download of the dataset when
  :func:`datasets.download_20newsgroups` is called. By `Manoj Kumar`_.

- Various fixes to the Gaussian processes subpackage by Vincent Dubourg
  and Jan Hendrik Metzen.

- Calling ``partial_fit`` with ``class_weight=='auto'`` throws an
  appropriate error message and suggests a work around.
  By :user:`Danny Sullivan <dsullivan7>`.

- :class:`RBFSampler <kernel_approximation.RBFSampler>` with ``gamma=g``
  formerly approximated :func:`rbf_kernel <metrics.pairwise.rbf_kernel>`
  with ``gamma=g/2.``; the definition of ``gamma`` is now consistent,
  which may substantially change your results if you use a fixed value.
  (If you cross-validated over ``gamma``, it probably doesn't matter
  too much.) By :user:`Dougal Sutherland <dougalsutherland>`.

- Pipeline object delegate the ``classes_`` attribute to the underlying
  estimator. It allows, for instance, to make bagging of a pipeline object.
  By `Arnaud Joly`_

- :class:`neighbors.NearestCentroid` now uses the median as the centroid
  when metric is set to ``manhattan``. It was using the mean before.
  By `Manoj Kumar`_

- Fix numerical stability issues in :class:`linear_model.SGDClassifier`
  and :class:`linear_model.SGDRegressor` by clipping large gradients and
  ensuring that weight decay rescaling is always positive (for large
  l2 regularization and large learning rate values).
  By `Olivier Grisel`_

- When `compute_full_tree` is set to "auto", the full tree is
  built when n_clusters is high and is early stopped when n_clusters is
  low, while the behavior should be vice-versa in
  :class:`cluster.AgglomerativeClustering` (and friends).
  This has been fixed By `Manoj Kumar`_

- Fix lazy centering of data in :func:`linear_model.enet_path` and
  :func:`linear_model.lasso_path`. It was centered around one. It has
  been changed to be centered around the origin. By `Manoj Kumar`_

- Fix handling of precomputed affinity matrices in
  :class:`cluster.AgglomerativeClustering` when using connectivity
  constraints. By :user:`Cathy Deng <cathydeng>`

- Correct ``partial_fit`` handling of ``class_prior`` for
  :class:`sklearn.naive_bayes.MultinomialNB` and
  :class:`sklearn.naive_bayes.BernoulliNB`. By `Trevor Stephens`_.

- Fixed a crash in :func:`metrics.precision_recall_fscore_support`
  when using unsorted ``labels`` in the multi-label setting.
  By `Andreas Müller`_.

- Avoid skipping the first nearest neighbor in the methods ``radius_neighbors``,
  ``kneighbors``, ``kneighbors_graph`` and ``radius_neighbors_graph`` in
  :class:`sklearn.neighbors.NearestNeighbors` and family, when the query
  data is not the same as fit data. By `Manoj Kumar`_.

- Fix log-density calculation in the :class:`mixture.GMM` with
  tied covariance. By `Will Dawson`_

- Fixed a scaling error in :class:`feature_selection.SelectFdr`
  where a factor ``n_features`` was missing. By `Andrew Tulloch`_

- Fix zero division in :class:`neighbors.KNeighborsRegressor` and related
  classes when using distance weighting and having identical data points.
  By `Garret-R <https://github.com/Garrett-R>`_.

- Fixed round off errors with non positive-definite covariance matrices
  in GMM. By :user:`Alexis Mignon <AlexisMignon>`.

- Fixed a error in the computation of conditional probabilities in
  :class:`naive_bayes.BernoulliNB`. By `Hanna Wallach`_.

- Make the method ``radius_neighbors`` of
  :class:`neighbors.NearestNeighbors` return the samples lying on the
  boundary for ``algorithm='brute'``. By `Yan Yi`_.

- Flip sign of ``dual_coef_`` of :class:`svm.SVC`
  to make it consistent with the documentation and
  ``decision_function``. By Artem Sobolev.

- Fixed handling of ties in :class:`isotonic.IsotonicRegression`.
  We now use the weighted average of targets (secondary method). By
  `Andreas Müller`_ and `Michael Bommarito <http://bommaritollc.com/>`_.

API changes summary
-------------------

- :class:`GridSearchCV <grid_search.GridSearchCV>` and
  :func:`cross_val_score <cross_validation.cross_val_score>` and other
  meta-estimators don't convert pandas DataFrames into arrays any more,
  allowing DataFrame specific operations in custom estimators.

- :func:`multiclass.fit_ovr`, :func:`multiclass.predict_ovr`,
  :func:`predict_proba_ovr`,
  :func:`multiclass.fit_ovo`, :func:`multiclass.predict_ovo`,
  :func:`multiclass.fit_ecoc` and :func:`multiclass.predict_ecoc`
  are deprecated. Use the underlying estimators instead.

- Nearest neighbors estimators used to take arbitrary keyword arguments
  and pass these to their distance metric. This will no longer be supported
  in scikit-learn 0.18; use the ``metric_params`` argument instead.

- `n_jobs` parameter of the fit method shifted to the constructor of the
       LinearRegression class.

- The ``predict_proba`` method of :class:`multiclass.OneVsRestClassifier`
  now returns two probabilities per sample in the multiclass case; this
  is consistent with other estimators and with the method's documentation,
  but previous versions accidentally returned only the positive
  probability. Fixed by Will Lamond and `Lars Buitinck`_.

- Change default value of precompute in :class:`ElasticNet` and :class:`Lasso`
  to False. Setting precompute to "auto" was found to be slower when
  n_samples > n_features since the computation of the Gram matrix is
  computationally expensive and outweighs the benefit of fitting the Gram
  for just one alpha.
  ``precompute="auto"`` is now deprecated and will be removed in 0.18
  By `Manoj Kumar`_.

- Expose ``positive`` option in :func:`linear_model.enet_path` and
  :func:`linear_model.enet_path` which constrains coefficients to be
  positive. By `Manoj Kumar`_.

- Users should now supply an explicit ``average`` parameter to
  :func:`sklearn.metrics.f1_score`, :func:`sklearn.metrics.fbeta_score`,
  :func:`sklearn.metrics.recall_score` and
  :func:`sklearn.metrics.precision_score` when performing multiclass
  or multilabel (i.e. not binary) classification. By `Joel Nothman`_.

- `scoring` parameter for cross validation now accepts `'f1_micro'`,
  `'f1_macro'` or `'f1_weighted'`. `'f1'` is now for binary classification
  only. Similar changes apply to `'precision'` and `'recall'`.
  By `Joel Nothman`_.

- The ``fit_intercept``, ``normalize`` and ``return_models`` parameters in
  :func:`linear_model.enet_path` and :func:`linear_model.lasso_path` have
  been removed. They were deprecated since 0.14

- From now onwards, all estimators will uniformly raise ``NotFittedError``
  (:class:`utils.validation.NotFittedError`), when any of the ``predict``
  like methods are called before the model is fit. By `Raghav RV`_.

- Input data validation was refactored for more consistent input
  validation. The ``check_arrays`` function was replaced by ``check_array``
  and ``check_X_y``. By `Andreas Müller`_.

- Allow ``X=None`` in the methods ``radius_neighbors``, ``kneighbors``,
  ``kneighbors_graph`` and ``radius_neighbors_graph`` in
  :class:`sklearn.neighbors.NearestNeighbors` and family. If set to None,
  then for every sample this avoids setting the sample itself as the
  first nearest neighbor. By `Manoj Kumar`_.

- Add parameter ``include_self`` in :func:`neighbors.kneighbors_graph`
  and :func:`neighbors.radius_neighbors_graph` which has to be explicitly
  set by the user. If set to True, then the sample itself is considered
  as the first nearest neighbor.

- `thresh` parameter is deprecated in favor of new `tol` parameter in
  :class:`GMM`, :class:`DPGMM` and :class:`VBGMM`. See `Enhancements`
  section for details. By `Hervé Bredin`_.

- Estimators will treat input with dtype object as numeric when possible.
  By `Andreas Müller`_

- Estimators now raise `ValueError` consistently when fitted on empty
  data (less than 1 sample or less than 1 feature for 2D input).
  By `Olivier Grisel`_.


- The ``shuffle`` option of :class:`.linear_model.SGDClassifier`,
  :class:`linear_model.SGDRegressor`, :class:`linear_model.Perceptron`,
  :class:`linear_model.PassiveAgressiveClassifier` and
  :class:`linear_model.PassiveAgressiveRegressor` now defaults to ``True``.

- :class:`cluster.DBSCAN` now uses a deterministic initialization. The
  `random_state` parameter is deprecated. By :user:`Erich Schubert <kno10>`.

Code Contributors
-----------------
A. Flaxman, Aaron Schumacher, Aaron Staple, abhishek thakur, Akshay, akshayah3,
Aldrian Obaja, Alexander Fabisch, Alexandre Gramfort, Alexis Mignon, Anders
Aagaard, Andreas Mueller, Andreas van Cranenburgh, Andrew Tulloch, Andrew
Walker, Antony Lee, Arnaud Joly, banilo, Barmaley.exe, Ben Davies, Benedikt
Koehler, bhsu, Boris Feld, Borja Ayerdi, Boyuan Deng, Brent Pedersen, Brian
Wignall, Brooke Osborn, Calvin Giles, Cathy Deng, Celeo, cgohlke, chebee7i,
Christian Stade-Schuldt, Christof Angermueller, Chyi-Kwei Yau, CJ Carey,
Clemens Brunner, Daiki Aminaka, Dan Blanchard, danfrankj, Danny Sullivan, David
Fletcher, Dmitrijs Milajevs, Dougal J. Sutherland, Erich Schubert, Fabian
Pedregosa, Florian Wilhelm, floydsoft, Félix-Antoine Fortin, Gael Varoquaux,
Garrett-R, Gilles Louppe, gpassino, gwulfs, Hampus Bengtsson, Hamzeh Alsalhi,
Hanna Wallach, Harry Mavroforakis, Hasil Sharma, Helder, Herve Bredin,
Hsiang-Fu Yu, Hugues SALAMIN, Ian Gilmore, Ilambharathi Kanniah, Imran Haque,
isms, Jake VanderPlas, Jan Dlabal, Jan Hendrik Metzen, Jatin Shah, Javier López
Peña, jdcaballero, Jean Kossaifi, Jeff Hammerbacher, Joel Nothman, Jonathan
Helmus, Joseph, Kaicheng Zhang, Kevin Markham, Kyle Beauchamp, Kyle Kastner,
Lagacherie Matthieu, Lars Buitinck, Laurent Direr, leepei, Loic Esteve, Luis
Pedro Coelho, Lukas Michelbacher, maheshakya, Manoj Kumar, Manuel, Mario
Michael Krell, Martin, Martin Billinger, Martin Ku, Mateusz Susik, Mathieu
Blondel, Matt Pico, Matt Terry, Matteo Visconti dOC, Matti Lyra, Max Linke,
Mehdi Cherti, Michael Bommarito, Michael Eickenberg, Michal Romaniuk, MLG,
mr.Shu, Nelle Varoquaux, Nicola Montecchio, Nicolas, Nikolay Mayorov, Noel
Dawe, Okal Billy, Olivier Grisel, Óscar Nájera, Paolo Puggioni, Peter
Prettenhofer, Pratap Vardhan, pvnguyen, queqichao, Rafael Carrascosa, Raghav R
V, Rahiel Kasim, Randall Mason, Rob Zinkov, Robert Bradshaw, Saket Choudhary,
Sam Nicholls, Samuel Charron, Saurabh Jha, sethdandridge, sinhrks, snuderl,
Stefan Otte, Stefan van der Walt, Steve Tjoa, swu, Sylvain Zimmer, tejesh95,
terrycojones, Thomas Delteil, Thomas Unterthiner, Tomas Kazmar, trevorstephens,
tttthomasssss, Tzu-Ming Kuo, ugurcaliskan, ugurthemaster, Vinayak Mehta,
Vincent Dubourg, Vjacheslav Murashkin, Vlad Niculae, wadawson, Wei Xue, Will
Lamond, Wu Jiang, x0l, Xinfan Meng, Yan Yi, Yu-Chin

.. _changes_0_15_2:

Version 0.15.2
==============

**September 4, 2014**

Bug fixes
---------

- Fixed handling of the ``p`` parameter of the Minkowski distance that was
  previously ignored in nearest neighbors models. By :user:`Nikolay
  Mayorov <nmayorov>`.

- Fixed duplicated alphas in :class:`linear_model.LassoLars` with early
  stopping on 32 bit Python. By `Olivier Grisel`_ and `Fabian Pedregosa`_.

- Fixed the build under Windows when scikit-learn is built with MSVC while
  NumPy is built with MinGW. By `Olivier Grisel`_ and :user:`Federico
  Vaggi <FedericoV>`.

- Fixed an array index overflow bug in the coordinate descent solver. By
  `Gael Varoquaux`_.

- Better handling of numpy 1.9 deprecation warnings. By `Gael Varoquaux`_.

- Removed unnecessary data copy in :class:`cluster.KMeans`.
  By `Gael Varoquaux`_.

- Explicitly close open files to avoid ``ResourceWarnings`` under Python 3.
  By Calvin Giles.

- The ``transform`` of :class:`discriminant_analysis.LinearDiscriminantAnalysis`
  now projects the input on the most discriminant directions. By Martin Billinger.

- Fixed potential overflow in ``_tree.safe_realloc`` by `Lars Buitinck`_.

- Performance optimization in :class:`isotonic.IsotonicRegression`.
  By Robert Bradshaw.

- ``nose`` is non-longer a runtime dependency to import ``sklearn``, only for
  running the tests. By `Joel Nothman`_.

- Many documentation and website fixes by `Joel Nothman`_, `Lars Buitinck`_
  :user:`Matt Pico <MattpSoftware>`, and others.

.. _changes_0_15_1:

Version 0.15.1
==============

**August 1, 2014**

Bug fixes
---------

- Made :func:`cross_validation.cross_val_score` use
  :class:`cross_validation.KFold` instead of
  :class:`cross_validation.StratifiedKFold` on multi-output classification
  problems. By :user:`Nikolay Mayorov <nmayorov>`.

- Support unseen labels :class:`preprocessing.LabelBinarizer` to restore
  the default behavior of 0.14.1 for backward compatibility. By
  :user:`Hamzeh Alsalhi <hamsal>`.

- Fixed the :class:`cluster.KMeans` stopping criterion that prevented early
  convergence detection. By Edward Raff and `Gael Varoquaux`_.

- Fixed the behavior of :class:`multiclass.OneVsOneClassifier`.
  in case of ties at the per-class vote level by computing the correct
  per-class sum of prediction scores. By `Andreas Müller`_.

- Made :func:`cross_validation.cross_val_score` and
  :class:`grid_search.GridSearchCV` accept Python lists as input data.
  This is especially useful for cross-validation and model selection of
  text processing pipelines. By `Andreas Müller`_.

- Fixed data input checks of most estimators to accept input data that
  implements the NumPy ``__array__`` protocol. This is the case for
  for ``pandas.Series`` and ``pandas.DataFrame`` in recent versions of
  pandas. By `Gael Varoquaux`_.

- Fixed a regression for :class:`linear_model.SGDClassifier` with
  ``class_weight="auto"`` on data with non-contiguous labels. By
  `Olivier Grisel`_.


.. _changes_0_15:

Version 0.15
============

**July 15, 2014**

Highlights
-----------

- Many speed and memory improvements all across the code

- Huge speed and memory improvements to random forests (and extra
  trees) that also benefit better from parallel computing.

- Incremental fit to :class:`BernoulliRBM <neural_network.BernoulliRBM>`

- Added :class:`cluster.AgglomerativeClustering` for hierarchical
  agglomerative clustering with average linkage, complete linkage and
  ward strategies.

- Added :class:`linear_model.RANSACRegressor` for robust regression
  models.

- Added dimensionality reduction with :class:`manifold.TSNE` which can be
  used to visualize high-dimensional data.


Changelog
---------

New features
............

- Added :class:`ensemble.BaggingClassifier` and
  :class:`ensemble.BaggingRegressor` meta-estimators for ensembling
  any kind of base estimator. See the :ref:`Bagging <bagging>` section of
  the user guide for details and examples. By `Gilles Louppe`_.

- New unsupervised feature selection algorithm
  :class:`feature_selection.VarianceThreshold`, by `Lars Buitinck`_.

- Added :class:`linear_model.RANSACRegressor` meta-estimator for the robust
  fitting of regression models. By :user:`Johannes Schönberger <ahojnnes>`.

- Added :class:`cluster.AgglomerativeClustering` for hierarchical
  agglomerative clustering with average linkage, complete linkage and
  ward strategies, by  `Nelle Varoquaux`_ and `Gael Varoquaux`_.

- Shorthand constructors :func:`pipeline.make_pipeline` and
  :func:`pipeline.make_union` were added by `Lars Buitinck`_.

- Shuffle option for :class:`cross_validation.StratifiedKFold`.
  By :user:`Jeffrey Blackburne <jblackburne>`.

- Incremental learning (``partial_fit``) for Gaussian Naive Bayes by
  Imran Haque.

- Added ``partial_fit`` to :class:`BernoulliRBM
  <neural_network.BernoulliRBM>`
  By :user:`Danny Sullivan <dsullivan7>`.

- Added :func:`learning_curve <learning_curve.learning_curve>` utility to
  chart performance with respect to training size. See
  :ref:`sphx_glr_auto_examples_model_selection_plot_learning_curve.py`. By Alexander Fabisch.

- Add positive option in :class:`LassoCV <linear_model.LassoCV>` and
  :class:`ElasticNetCV <linear_model.ElasticNetCV>`.
  By Brian Wignall and `Alexandre Gramfort`_.

- Added :class:`linear_model.MultiTaskElasticNetCV` and
  :class:`linear_model.MultiTaskLassoCV`. By `Manoj Kumar`_.

- Added :class:`manifold.TSNE`. By Alexander Fabisch.

Enhancements
............

- Add sparse input support to :class:`ensemble.AdaBoostClassifier` and
  :class:`ensemble.AdaBoostRegressor` meta-estimators.
  By :user:`Hamzeh Alsalhi <hamsal>`.

- Memory improvements of decision trees, by `Arnaud Joly`_.

- Decision trees can now be built in best-first manner by using ``max_leaf_nodes``
  as the stopping criteria. Refactored the tree code to use either a
  stack or a priority queue for tree building.
  By `Peter Prettenhofer`_ and `Gilles Louppe`_.

- Decision trees can now be fitted on fortran- and c-style arrays, and
  non-continuous arrays without the need to make a copy.
  If the input array has a different dtype than ``np.float32``, a fortran-
  style copy will be made since fortran-style memory layout has speed
  advantages. By `Peter Prettenhofer`_ and `Gilles Louppe`_.

- Speed improvement of regression trees by optimizing the
  the computation of the mean square error criterion. This lead
  to speed improvement of the tree, forest and gradient boosting tree
  modules. By `Arnaud Joly`_

- The ``img_to_graph`` and ``grid_tograph`` functions in
  :mod:`sklearn.feature_extraction.image` now return ``np.ndarray``
  instead of ``np.matrix`` when ``return_as=np.ndarray``.  See the
  Notes section for more information on compatibility.

- Changed the internal storage of decision trees to use a struct array.
  This fixed some small bugs, while improving code and providing a small
  speed gain. By `Joel Nothman`_.

- Reduce memory usage and overhead when fitting and predicting with forests
  of randomized trees in parallel with ``n_jobs != 1`` by leveraging new
  threading backend of joblib 0.8 and releasing the GIL in the tree fitting
  Cython code.  By `Olivier Grisel`_ and `Gilles Louppe`_.

- Speed improvement of the :mod:`sklearn.ensemble.gradient_boosting` module.
  By `Gilles Louppe`_ and `Peter Prettenhofer`_.

- Various enhancements to the  :mod:`sklearn.ensemble.gradient_boosting`
  module: a ``warm_start`` argument to fit additional trees,
  a ``max_leaf_nodes`` argument to fit GBM style trees,
  a ``monitor`` fit argument to inspect the estimator during training, and
  refactoring of the verbose code. By `Peter Prettenhofer`_.

- Faster :class:`sklearn.ensemble.ExtraTrees` by caching feature values.
  By `Arnaud Joly`_.

- Faster depth-based tree building algorithm such as decision tree,
  random forest, extra trees or gradient tree boosting (with depth based
  growing strategy) by avoiding trying to split on found constant features
  in the sample subset. By `Arnaud Joly`_.

- Add ``min_weight_fraction_leaf`` pre-pruning parameter to tree-based
  methods: the minimum weighted fraction of the input samples required to be
  at a leaf node. By `Noel Dawe`_.

- Added :func:`metrics.pairwise_distances_argmin_min`, by Philippe Gervais.

- Added predict method to :class:`cluster.AffinityPropagation` and
  :class:`cluster.MeanShift`, by `Mathieu Blondel`_.

- Vector and matrix multiplications have been optimised throughout the
  library by `Denis Engemann`_, and `Alexandre Gramfort`_.
  In particular, they should take less memory with older NumPy versions
  (prior to 1.7.2).

- Precision-recall and ROC examples now use train_test_split, and have more
  explanation of why these metrics are useful. By `Kyle Kastner`_

- The training algorithm for :class:`decomposition.NMF` is faster for
  sparse matrices and has much lower memory complexity, meaning it will
  scale up gracefully to large datasets. By `Lars Buitinck`_.

- Added svd_method option with default value to "randomized" to
  :class:`decomposition.FactorAnalysis` to save memory and
  significantly speedup computation by `Denis Engemann`_, and
  `Alexandre Gramfort`_.

- Changed :class:`cross_validation.StratifiedKFold` to try and
  preserve as much of the original ordering of samples as possible so as
  not to hide overfitting on datasets with a non-negligible level of
  samples dependency.
  By `Daniel Nouri`_ and `Olivier Grisel`_.

- Add multi-output support to :class:`gaussian_process.GaussianProcess`
  by John Novak.

- Support for precomputed distance matrices in nearest neighbor estimators
  by `Robert Layton`_ and `Joel Nothman`_.

- Norm computations optimized for NumPy 1.6 and later versions by
  `Lars Buitinck`_. In particular, the k-means algorithm no longer
  needs a temporary data structure the size of its input.

- :class:`dummy.DummyClassifier` can now be used to predict a constant
  output value. By `Manoj Kumar`_.

- :class:`dummy.DummyRegressor` has now a strategy parameter which allows
  to predict the mean, the median of the training set or a constant
  output value. By :user:`Maheshakya Wijewardena <maheshakya>`.

- Multi-label classification output in multilabel indicator format
  is now supported by :func:`metrics.roc_auc_score` and
  :func:`metrics.average_precision_score` by `Arnaud Joly`_.

- Significant performance improvements (more than 100x speedup for
  large problems) in :class:`isotonic.IsotonicRegression` by
  `Andrew Tulloch`_.

- Speed and memory usage improvements to the SGD algorithm for linear
  models: it now uses threads, not separate processes, when ``n_jobs>1``.
  By `Lars Buitinck`_.

- Grid search and cross validation allow NaNs in the input arrays so that
  preprocessors such as :class:`preprocessing.Imputer
  <preprocessing.Imputer>` can be trained within the cross validation loop,
  avoiding potentially skewed results.

- Ridge regression can now deal with sample weights in feature space
  (only sample space until then). By :user:`Michael Eickenberg <eickenberg>`.
  Both solutions are provided by the Cholesky solver.

- Several classification and regression metrics now support weighted
  samples with the new ``sample_weight`` argument:
  :func:`metrics.accuracy_score`,
  :func:`metrics.zero_one_loss`,
  :func:`metrics.precision_score`,
  :func:`metrics.average_precision_score`,
  :func:`metrics.f1_score`,
  :func:`metrics.fbeta_score`,
  :func:`metrics.recall_score`,
  :func:`metrics.roc_auc_score`,
  :func:`metrics.explained_variance_score`,
  :func:`metrics.mean_squared_error`,
  :func:`metrics.mean_absolute_error`,
  :func:`metrics.r2_score`.
  By `Noel Dawe`_.

- Speed up of the sample generator
  :func:`datasets.make_multilabel_classification`. By `Joel Nothman`_.

Documentation improvements
...........................

- The :ref:`Working With Text Data <text_data_tutorial>` tutorial
  has now been worked in to the main documentation's tutorial section.
  Includes exercises and skeletons for tutorial presentation.
  Original tutorial created by several authors including
  `Olivier Grisel`_, Lars Buitinck and many others.
  Tutorial integration into the scikit-learn documentation
  by `Jaques Grobler`_

- Added :ref:`Computational Performance <computational_performance>`
  documentation. Discussion and examples of prediction latency / throughput
  and different factors that have influence over speed. Additional tips for
  building faster models and choosing a relevant compromise between speed
  and predictive power.
  By :user:`Eustache Diemert <oddskool>`.

Bug fixes
.........

- Fixed bug in :class:`decomposition.MiniBatchDictionaryLearning` :
  ``partial_fit`` was not working properly.

- Fixed bug in :class:`linear_model.stochastic_gradient` :
  ``l1_ratio`` was used as ``(1.0 - l1_ratio)`` .

- Fixed bug in :class:`multiclass.OneVsOneClassifier` with string
  labels

- Fixed a bug in :class:`LassoCV <linear_model.LassoCV>` and
  :class:`ElasticNetCV <linear_model.ElasticNetCV>`: they would not
  pre-compute the Gram matrix with ``precompute=True`` or
  ``precompute="auto"`` and ``n_samples > n_features``. By `Manoj Kumar`_.

- Fixed incorrect estimation of the degrees of freedom in
  :func:`feature_selection.f_regression` when variates are not centered.
  By :user:`Virgile Fritsch <VirgileFritsch>`.

- Fixed a race condition in parallel processing with
  ``pre_dispatch != "all"`` (for instance, in ``cross_val_score``).
  By `Olivier Grisel`_.

- Raise error in :class:`cluster.FeatureAgglomeration` and
  :class:`cluster.WardAgglomeration` when no samples are given,
  rather than returning meaningless clustering.

- Fixed bug in :class:`gradient_boosting.GradientBoostingRegressor` with
  ``loss='huber'``: ``gamma`` might have not been initialized.

- Fixed feature importances as computed with a forest of randomized trees
  when fit with ``sample_weight != None`` and/or with ``bootstrap=True``.
  By `Gilles Louppe`_.

API changes summary
-------------------

- :mod:`sklearn.hmm` is deprecated. Its removal is planned
  for the 0.17 release.

- Use of :class:`covariance.EllipticEnvelop` has now been removed after
  deprecation.
  Please use :class:`covariance.EllipticEnvelope` instead.

- :class:`cluster.Ward` is deprecated. Use
  :class:`cluster.AgglomerativeClustering` instead.

- :class:`cluster.WardClustering` is deprecated. Use
- :class:`cluster.AgglomerativeClustering` instead.

- :class:`cross_validation.Bootstrap` is deprecated.
  :class:`cross_validation.KFold` or
  :class:`cross_validation.ShuffleSplit` are recommended instead.

- Direct support for the sequence of sequences (or list of lists) multilabel
  format is deprecated. To convert to and from the supported binary
  indicator matrix format, use
  :class:`MultiLabelBinarizer <preprocessing.MultiLabelBinarizer>`.
  By `Joel Nothman`_.

- Add score method to :class:`PCA <decomposition.PCA>` following the model of
  probabilistic PCA and deprecate
  :class:`ProbabilisticPCA <decomposition.ProbabilisticPCA>` model whose
  score implementation is not correct. The computation now also exploits the
  matrix inversion lemma for faster computation. By `Alexandre Gramfort`_.

- The score method of :class:`FactorAnalysis <decomposition.FactorAnalysis>`
  now returns the average log-likelihood of the samples. Use score_samples
  to get log-likelihood of each sample. By `Alexandre Gramfort`_.

- Generating boolean masks (the setting ``indices=False``)
  from cross-validation generators is deprecated.
  Support for masks will be removed in 0.17.
  The generators have produced arrays of indices by default since 0.10.
  By `Joel Nothman`_.

- 1-d arrays containing strings with ``dtype=object`` (as used in Pandas)
  are now considered valid classification targets. This fixes a regression
  from version 0.13 in some classifiers. By `Joel Nothman`_.

- Fix wrong ``explained_variance_ratio_`` attribute in
  :class:`RandomizedPCA <decomposition.RandomizedPCA>`.
  By `Alexandre Gramfort`_.

- Fit alphas for each ``l1_ratio`` instead of ``mean_l1_ratio`` in
  :class:`linear_model.ElasticNetCV` and :class:`linear_model.LassoCV`.
  This changes the shape of ``alphas_`` from ``(n_alphas,)`` to
  ``(n_l1_ratio, n_alphas)`` if the ``l1_ratio`` provided is a 1-D array like
  object of length greater than one.
  By `Manoj Kumar`_.

- Fix :class:`linear_model.ElasticNetCV` and :class:`linear_model.LassoCV`
  when fitting intercept and input data is sparse. The automatic grid
  of alphas was not computed correctly and the scaling with normalize
  was wrong. By `Manoj Kumar`_.

- Fix wrong maximal number of features drawn (``max_features``) at each split
  for decision trees, random forests and gradient tree boosting.
  Previously, the count for the number of drawn features started only after
  one non constant features in the split. This bug fix will affect
  computational and generalization performance of those algorithms in the
  presence of constant features. To get back previous generalization
  performance, you should modify the value of ``max_features``.
  By `Arnaud Joly`_.

- Fix wrong maximal number of features drawn (``max_features``) at each split
  for :class:`ensemble.ExtraTreesClassifier` and
  :class:`ensemble.ExtraTreesRegressor`. Previously, only non constant
  features in the split was counted as drawn. Now constant features are
  counted as drawn. Furthermore at least one feature must be non constant
  in order to make a valid split. This bug fix will affect
  computational and generalization performance of extra trees in the
  presence of constant features. To get back previous generalization
  performance, you should modify the value of ``max_features``.
  By `Arnaud Joly`_.

- Fix :func:`utils.compute_class_weight` when ``class_weight=="auto"``.
  Previously it was broken for input of non-integer ``dtype`` and the
  weighted array that was returned was wrong. By `Manoj Kumar`_.

- Fix :class:`cross_validation.Bootstrap` to return ``ValueError``
  when ``n_train + n_test > n``. By :user:`Ronald Phlypo <rphlypo>`.


People
------

List of contributors for release 0.15 by number of commits.

* 312	Olivier Grisel
* 275	Lars Buitinck
* 221	Gael Varoquaux
* 148	Arnaud Joly
* 134	Johannes Schönberger
* 119	Gilles Louppe
* 113	Joel Nothman
* 111	Alexandre Gramfort
*  95	Jaques Grobler
*  89	Denis Engemann
*  83	Peter Prettenhofer
*  83	Alexander Fabisch
*  62	Mathieu Blondel
*  60	Eustache Diemert
*  60	Nelle Varoquaux
*  49	Michael Bommarito
*  45	Manoj-Kumar-S
*  28	Kyle Kastner
*  26	Andreas Mueller
*  22	Noel Dawe
*  21	Maheshakya Wijewardena
*  21	Brooke Osborn
*  21	Hamzeh Alsalhi
*  21	Jake VanderPlas
*  21	Philippe Gervais
*  19	Bala Subrahmanyam Varanasi
*  12	Ronald Phlypo
*  10	Mikhail Korobov
*   8	Thomas Unterthiner
*   8	Jeffrey Blackburne
*   8	eltermann
*   8	bwignall
*   7	Ankit Agrawal
*   7	CJ Carey
*   6	Daniel Nouri
*   6	Chen Liu
*   6	Michael Eickenberg
*   6	ugurthemaster
*   5	Aaron Schumacher
*   5	Baptiste Lagarde
*   5	Rajat Khanduja
*   5	Robert McGibbon
*   5	Sergio Pascual
*   4	Alexis Metaireau
*   4	Ignacio Rossi
*   4	Virgile Fritsch
*   4	Sebastian Säger
*   4	Ilambharathi Kanniah
*   4	sdenton4
*   4	Robert Layton
*   4	Alyssa
*   4	Amos Waterland
*   3	Andrew Tulloch
*   3	murad
*   3	Steven Maude
*   3	Karol Pysniak
*   3	Jacques Kvam
*   3	cgohlke
*   3	cjlin
*   3	Michael Becker
*   3	hamzeh
*   3	Eric Jacobsen
*   3	john collins
*   3	kaushik94
*   3	Erwin Marsi
*   2	csytracy
*   2	LK
*   2	Vlad Niculae
*   2	Laurent Direr
*   2	Erik Shilts
*   2	Raul Garreta
*   2	Yoshiki Vázquez Baeza
*   2	Yung Siang Liau
*   2	abhishek thakur
*   2	James Yu
*   2	Rohit Sivaprasad
*   2	Roland Szabo
*   2	amormachine
*   2	Alexis Mignon
*   2	Oscar Carlsson
*   2	Nantas Nardelli
*   2	jess010
*   2	kowalski87
*   2	Andrew Clegg
*   2	Federico Vaggi
*   2	Simon Frid
*   2	Félix-Antoine Fortin
*   1	Ralf Gommers
*   1	t-aft
*   1	Ronan Amicel
*   1	Rupesh Kumar Srivastava
*   1	Ryan Wang
*   1	Samuel Charron
*   1	Samuel St-Jean
*   1	Fabian Pedregosa
*   1	Skipper Seabold
*   1	Stefan Walk
*   1	Stefan van der Walt
*   1	Stephan Hoyer
*   1	Allen Riddell
*   1	Valentin Haenel
*   1	Vijay Ramesh
*   1	Will Myers
*   1	Yaroslav Halchenko
*   1	Yoni Ben-Meshulam
*   1	Yury V. Zaytsev
*   1	adrinjalali
*   1	ai8rahim
*   1	alemagnani
*   1	alex
*   1	benjamin wilson
*   1	chalmerlowe
*   1	dzikie drożdże
*   1	jamestwebber
*   1	matrixorz
*   1	popo
*   1	samuela
*   1	François Boulogne
*   1	Alexander Measure
*   1	Ethan White
*   1	Guilherme Trein
*   1	Hendrik Heuer
*   1	IvicaJovic
*   1	Jan Hendrik Metzen
*   1	Jean Michel Rouly
*   1	Eduardo Ariño de la Rubia
*   1	Jelle Zijlstra
*   1	Eddy L O Jansson
*   1	Denis
*   1	John
*   1	John Schmidt
*   1	Jorge Cañardo Alastuey
*   1	Joseph Perla
*   1	Joshua Vredevoogd
*   1	José Ricardo
*   1	Julien Miotte
*   1	Kemal Eren
*   1	Kenta Sato
*   1	David Cournapeau
*   1	Kyle Kelley
*   1	Daniele Medri
*   1	Laurent Luce
*   1	Laurent Pierron
*   1	Luis Pedro Coelho
*   1	DanielWeitzenfeld
*   1	Craig Thompson
*   1	Chyi-Kwei Yau
*   1	Matthew Brett
*   1	Matthias Feurer
*   1	Max Linke
*   1	Chris Filo Gorgolewski
*   1	Charles Earl
*   1	Michael Hanke
*   1	Michele Orrù
*   1	Bryan Lunt
*   1	Brian Kearns
*   1	Paul Butler
*   1	Paweł Mandera
*   1	Peter
*   1	Andrew Ash
*   1	Pietro Zambelli
*   1	staubda


.. _changes_0_14:

Version 0.14
===============

**August 7, 2013**

Changelog
---------

- Missing values with sparse and dense matrices can be imputed with the
  transformer :class:`preprocessing.Imputer` by `Nicolas Trésegnie`_.

- The core implementation of decisions trees has been rewritten from
  scratch, allowing for faster tree induction and lower memory
  consumption in all tree-based estimators. By `Gilles Louppe`_.

- Added :class:`ensemble.AdaBoostClassifier` and
  :class:`ensemble.AdaBoostRegressor`, by `Noel Dawe`_  and
  `Gilles Louppe`_. See the :ref:`AdaBoost <adaboost>` section of the user
  guide for details and examples.

- Added :class:`grid_search.RandomizedSearchCV` and
  :class:`grid_search.ParameterSampler` for randomized hyperparameter
  optimization. By `Andreas Müller`_.

- Added :ref:`biclustering <biclustering>` algorithms
  (:class:`sklearn.cluster.bicluster.SpectralCoclustering` and
  :class:`sklearn.cluster.bicluster.SpectralBiclustering`), data
  generation methods (:func:`sklearn.datasets.make_biclusters` and
  :func:`sklearn.datasets.make_checkerboard`), and scoring metrics
  (:func:`sklearn.metrics.consensus_score`). By `Kemal Eren`_.

- Added :ref:`Restricted Boltzmann Machines<rbm>`
  (:class:`neural_network.BernoulliRBM`). By `Yann Dauphin`_.

- Python 3 support by :user:`Justin Vincent <justinvf>`, `Lars Buitinck`_,
  :user:`Subhodeep Moitra <smoitra87>` and `Olivier Grisel`_. All tests now pass under
  Python 3.3.

- Ability to pass one penalty (alpha value) per target in
  :class:`linear_model.Ridge`, by @eickenberg and `Mathieu Blondel`_.

- Fixed :mod:`sklearn.linear_model.stochastic_gradient.py` L2 regularization
  issue (minor practical significance).
  By :user:`Norbert Crombach <norbert>` and `Mathieu Blondel`_ .

- Added an interactive version of `Andreas Müller`_'s
  `Machine Learning Cheat Sheet (for scikit-learn)
  <http://peekaboo-vision.blogspot.de/2013/01/machine-learning-cheat-sheet-for-scikit.html>`_
  to the documentation. See :ref:`Choosing the right estimator <ml_map>`.
  By `Jaques Grobler`_.

- :class:`grid_search.GridSearchCV` and
  :func:`cross_validation.cross_val_score` now support the use of advanced
  scoring function such as area under the ROC curve and f-beta scores.
  See :ref:`scoring_parameter` for details. By `Andreas Müller`_
  and `Lars Buitinck`_.
  Passing a function from :mod:`sklearn.metrics` as ``score_func`` is
  deprecated.

- Multi-label classification output is now supported by
  :func:`metrics.accuracy_score`, :func:`metrics.zero_one_loss`,
  :func:`metrics.f1_score`, :func:`metrics.fbeta_score`,
  :func:`metrics.classification_report`,
  :func:`metrics.precision_score` and :func:`metrics.recall_score`
  by `Arnaud Joly`_.

- Two new metrics :func:`metrics.hamming_loss` and
  :func:`metrics.jaccard_similarity_score`
  are added with multi-label support by `Arnaud Joly`_.

- Speed and memory usage improvements in
  :class:`feature_extraction.text.CountVectorizer` and
  :class:`feature_extraction.text.TfidfVectorizer`,
  by Jochen Wersdörfer and Roman Sinayev.

- The ``min_df`` parameter in
  :class:`feature_extraction.text.CountVectorizer` and
  :class:`feature_extraction.text.TfidfVectorizer`, which used to be 2,
  has been reset to 1 to avoid unpleasant surprises (empty vocabularies)
  for novice users who try it out on tiny document collections.
  A value of at least 2 is still recommended for practical use.

- :class:`svm.LinearSVC`, :class:`linear_model.SGDClassifier` and
  :class:`linear_model.SGDRegressor` now have a ``sparsify`` method that
  converts their ``coef_`` into a sparse matrix, meaning stored models
  trained using these estimators can be made much more compact.

- :class:`linear_model.SGDClassifier` now produces multiclass probability
  estimates when trained under log loss or modified Huber loss.

- Hyperlinks to documentation in example code on the website by
  :user:`Martin Luessi <mluessi>`.

- Fixed bug in :class:`preprocessing.MinMaxScaler` causing incorrect scaling
  of the features for non-default ``feature_range`` settings. By `Andreas
  Müller`_.

- ``max_features`` in :class:`tree.DecisionTreeClassifier`,
  :class:`tree.DecisionTreeRegressor` and all derived ensemble estimators
  now supports percentage values. By `Gilles Louppe`_.

- Performance improvements in :class:`isotonic.IsotonicRegression` by
  `Nelle Varoquaux`_.

- :func:`metrics.accuracy_score` has an option normalize to return
  the fraction or the number of correctly classified sample
  by `Arnaud Joly`_.

- Added :func:`metrics.log_loss` that computes log loss, aka cross-entropy
  loss. By Jochen Wersdörfer and `Lars Buitinck`_.

- A bug that caused :class:`ensemble.AdaBoostClassifier`'s to output
  incorrect probabilities has been fixed.

- Feature selectors now share a mixin providing consistent ``transform``,
  ``inverse_transform`` and ``get_support`` methods. By `Joel Nothman`_.

- A fitted :class:`grid_search.GridSearchCV` or
  :class:`grid_search.RandomizedSearchCV` can now generally be pickled.
  By `Joel Nothman`_.

- Refactored and vectorized implementation of :func:`metrics.roc_curve`
  and :func:`metrics.precision_recall_curve`. By `Joel Nothman`_.

- The new estimator :class:`sklearn.decomposition.TruncatedSVD`
  performs dimensionality reduction using SVD on sparse matrices,
  and can be used for latent semantic analysis (LSA).
  By `Lars Buitinck`_.

- Added self-contained example of out-of-core learning on text data
  :ref:`sphx_glr_auto_examples_applications_plot_out_of_core_classification.py`.
  By :user:`Eustache Diemert <oddskool>`.

- The default number of components for
  :class:`sklearn.decomposition.RandomizedPCA` is now correctly documented
  to be ``n_features``. This was the default behavior, so programs using it
  will continue to work as they did.

- :class:`sklearn.cluster.KMeans` now fits several orders of magnitude
  faster on sparse data (the speedup depends on the sparsity). By
  `Lars Buitinck`_.

- Reduce memory footprint of FastICA by `Denis Engemann`_ and
  `Alexandre Gramfort`_.

- Verbose output in :mod:`sklearn.ensemble.gradient_boosting` now uses
  a column format and prints progress in decreasing frequency.
  It also shows the remaining time. By `Peter Prettenhofer`_.

- :mod:`sklearn.ensemble.gradient_boosting` provides out-of-bag improvement
  :attr:`~sklearn.ensemble.GradientBoostingRegressor.oob_improvement_`
  rather than the OOB score for model selection. An example that shows
  how to use OOB estimates to select the number of trees was added.
  By `Peter Prettenhofer`_.

- Most metrics now support string labels for multiclass classification
  by `Arnaud Joly`_ and `Lars Buitinck`_.

- New OrthogonalMatchingPursuitCV class by `Alexandre Gramfort`_
  and `Vlad Niculae`_.

- Fixed a bug in :class:`sklearn.covariance.GraphLassoCV`: the
  'alphas' parameter now works as expected when given a list of
  values. By Philippe Gervais.

- Fixed an important bug in :class:`sklearn.covariance.GraphLassoCV`
  that prevented all folds provided by a CV object to be used (only
  the first 3 were used). When providing a CV object, execution
  time may thus increase significantly compared to the previous
  version (bug results are correct now). By Philippe Gervais.

- :class:`cross_validation.cross_val_score` and the :mod:`grid_search`
  module is now tested with multi-output data by `Arnaud Joly`_.

- :func:`datasets.make_multilabel_classification` can now return
  the output in label indicator multilabel format  by `Arnaud Joly`_.

- K-nearest neighbors, :class:`neighbors.KNeighborsRegressor`
  and :class:`neighbors.RadiusNeighborsRegressor`,
  and radius neighbors, :class:`neighbors.RadiusNeighborsRegressor` and
  :class:`neighbors.RadiusNeighborsClassifier` support multioutput data
  by `Arnaud Joly`_.

- Random state in LibSVM-based estimators (:class:`svm.SVC`, :class:`NuSVC`,
  :class:`OneClassSVM`, :class:`svm.SVR`, :class:`svm.NuSVR`) can now be
  controlled.  This is useful to ensure consistency in the probability
  estimates for the classifiers trained with ``probability=True``. By
  `Vlad Niculae`_.

- Out-of-core learning support for discrete naive Bayes classifiers
  :class:`sklearn.naive_bayes.MultinomialNB` and
  :class:`sklearn.naive_bayes.BernoulliNB` by adding the ``partial_fit``
  method by `Olivier Grisel`_.

- New website design and navigation by `Gilles Louppe`_, `Nelle Varoquaux`_,
  Vincent Michel and `Andreas Müller`_.

- Improved documentation on :ref:`multi-class, multi-label and multi-output
  classification <multiclass>` by `Yannick Schwartz`_ and `Arnaud Joly`_.

- Better input and error handling in the :mod:`metrics` module by
  `Arnaud Joly`_ and `Joel Nothman`_.

- Speed optimization of the :mod:`hmm` module by :user:`Mikhail Korobov <kmike>`

- Significant speed improvements for :class:`sklearn.cluster.DBSCAN`
  by `cleverless <https://github.com/cleverless>`_


API changes summary
-------------------

- The :func:`auc_score` was renamed :func:`roc_auc_score`.

- Testing scikit-learn with ``sklearn.test()`` is deprecated. Use
  ``nosetests sklearn`` from the command line.

- Feature importances in :class:`tree.DecisionTreeClassifier`,
  :class:`tree.DecisionTreeRegressor` and all derived ensemble estimators
  are now computed on the fly when accessing  the ``feature_importances_``
  attribute. Setting ``compute_importances=True`` is no longer required.
  By `Gilles Louppe`_.

- :class:`linear_model.lasso_path` and
  :class:`linear_model.enet_path` can return its results in the same
  format as that of :class:`linear_model.lars_path`. This is done by
  setting the ``return_models`` parameter to ``False``. By
  `Jaques Grobler`_ and `Alexandre Gramfort`_

- :class:`grid_search.IterGrid` was renamed to
  :class:`grid_search.ParameterGrid`.

- Fixed bug in :class:`KFold` causing imperfect class balance in some
  cases. By `Alexandre Gramfort`_ and Tadej Janež.

- :class:`sklearn.neighbors.BallTree` has been refactored, and a
  :class:`sklearn.neighbors.KDTree` has been
  added which shares the same interface.  The Ball Tree now works with
  a wide variety of distance metrics.  Both classes have many new
  methods, including single-tree and dual-tree queries, breadth-first
  and depth-first searching, and more advanced queries such as
  kernel density estimation and 2-point correlation functions.
  By `Jake Vanderplas`_

- Support for scipy.spatial.cKDTree within neighbors queries has been
  removed, and the functionality replaced with the new :class:`KDTree`
  class.

- :class:`sklearn.neighbors.KernelDensity` has been added, which performs
  efficient kernel density estimation with a variety of kernels.

- :class:`sklearn.decomposition.KernelPCA` now always returns output with
  ``n_components`` components, unless the new parameter ``remove_zero_eig``
  is set to ``True``. This new behavior is consistent with the way
  kernel PCA was always documented; previously, the removal of components
  with zero eigenvalues was tacitly performed on all data.

- ``gcv_mode="auto"`` no longer tries to perform SVD on a densified
  sparse matrix in :class:`sklearn.linear_model.RidgeCV`.

- Sparse matrix support in :class:`sklearn.decomposition.RandomizedPCA`
  is now deprecated in favor of the new ``TruncatedSVD``.

- :class:`cross_validation.KFold` and
  :class:`cross_validation.StratifiedKFold` now enforce `n_folds >= 2`
  otherwise a ``ValueError`` is raised. By `Olivier Grisel`_.

- :func:`datasets.load_files`'s ``charset`` and ``charset_errors``
  parameters were renamed ``encoding`` and ``decode_errors``.

- Attribute ``oob_score_`` in :class:`sklearn.ensemble.GradientBoostingRegressor`
  and :class:`sklearn.ensemble.GradientBoostingClassifier`
  is deprecated and has been replaced by ``oob_improvement_`` .

- Attributes in OrthogonalMatchingPursuit have been deprecated
  (copy_X, Gram, ...) and precompute_gram renamed precompute
  for consistency. See #2224.

- :class:`sklearn.preprocessing.StandardScaler` now converts integer input
  to float, and raises a warning. Previously it rounded for dense integer
  input.

- :class:`sklearn.multiclass.OneVsRestClassifier` now has a
  ``decision_function`` method. This will return the distance of each
  sample from the decision boundary for each class, as long as the
  underlying estimators implement the ``decision_function`` method.
  By `Kyle Kastner`_.

- Better input validation, warning on unexpected shapes for y.

People
------
List of contributors for release 0.14 by number of commits.

 * 277  Gilles Louppe
 * 245  Lars Buitinck
 * 187  Andreas Mueller
 * 124  Arnaud Joly
 * 112  Jaques Grobler
 * 109  Gael Varoquaux
 * 107  Olivier Grisel
 * 102  Noel Dawe
 *  99  Kemal Eren
 *  79  Joel Nothman
 *  75  Jake VanderPlas
 *  73  Nelle Varoquaux
 *  71  Vlad Niculae
 *  65  Peter Prettenhofer
 *  64  Alexandre Gramfort
 *  54  Mathieu Blondel
 *  38  Nicolas Trésegnie
 *  35  eustache
 *  27  Denis Engemann
 *  25  Yann N. Dauphin
 *  19  Justin Vincent
 *  17  Robert Layton
 *  15  Doug Coleman
 *  14  Michael Eickenberg
 *  13  Robert Marchman
 *  11  Fabian Pedregosa
 *  11  Philippe Gervais
 *  10  Jim Holmström
 *  10  Tadej Janež
 *  10  syhw
 *   9  Mikhail Korobov
 *   9  Steven De Gryze
 *   8  sergeyf
 *   7  Ben Root
 *   7  Hrishikesh Huilgolkar
 *   6  Kyle Kastner
 *   6  Martin Luessi
 *   6  Rob Speer
 *   5  Federico Vaggi
 *   5  Raul Garreta
 *   5  Rob Zinkov
 *   4  Ken Geis
 *   3  A. Flaxman
 *   3  Denton Cockburn
 *   3  Dougal Sutherland
 *   3  Ian Ozsvald
 *   3  Johannes Schönberger
 *   3  Robert McGibbon
 *   3  Roman Sinayev
 *   3  Szabo Roland
 *   2  Diego Molla
 *   2  Imran Haque
 *   2  Jochen Wersdörfer
 *   2  Sergey Karayev
 *   2  Yannick Schwartz
 *   2  jamestwebber
 *   1  Abhijeet Kolhe
 *   1  Alexander Fabisch
 *   1  Bastiaan van den Berg
 *   1  Benjamin Peterson
 *   1  Daniel Velkov
 *   1  Fazlul Shahriar
 *   1  Felix Brockherde
 *   1  Félix-Antoine Fortin
 *   1  Harikrishnan S
 *   1  Jack Hale
 *   1  JakeMick
 *   1  James McDermott
 *   1  John Benediktsson
 *   1  John Zwinck
 *   1  Joshua Vredevoogd
 *   1  Justin Pati
 *   1  Kevin Hughes
 *   1  Kyle Kelley
 *   1  Matthias Ekman
 *   1  Miroslav Shubernetskiy
 *   1  Naoki Orii
 *   1  Norbert Crombach
 *   1  Rafael Cunha de Almeida
 *   1  Rolando Espinoza La fuente
 *   1  Seamus Abshere
 *   1  Sergey Feldman
 *   1  Sergio Medina
 *   1  Stefano Lattarini
 *   1  Steve Koch
 *   1  Sturla Molden
 *   1  Thomas Jarosch
 *   1  Yaroslav Halchenko

.. _changes_0_13_1:

Version 0.13.1
==============

**February 23, 2013**

The 0.13.1 release only fixes some bugs and does not add any new functionality.

Changelog
---------

- Fixed a testing error caused by the function :func:`cross_validation.train_test_split` being
  interpreted as a test by `Yaroslav Halchenko`_.

- Fixed a bug in the reassignment of small clusters in the :class:`cluster.MiniBatchKMeans`
  by `Gael Varoquaux`_.

- Fixed default value of ``gamma`` in :class:`decomposition.KernelPCA` by `Lars Buitinck`_.

- Updated joblib to ``0.7.0d`` by `Gael Varoquaux`_.

- Fixed scaling of the deviance in :class:`ensemble.GradientBoostingClassifier` by `Peter Prettenhofer`_.

- Better tie-breaking in :class:`multiclass.OneVsOneClassifier` by `Andreas Müller`_.

- Other small improvements to tests and documentation.

People
------
List of contributors for release 0.13.1 by number of commits.
 * 16  `Lars Buitinck`_
 * 12  `Andreas Müller`_
 *  8  `Gael Varoquaux`_
 *  5  Robert Marchman
 *  3  `Peter Prettenhofer`_
 *  2  Hrishikesh Huilgolkar
 *  1  Bastiaan van den Berg
 *  1  Diego Molla
 *  1  `Gilles Louppe`_
 *  1  `Mathieu Blondel`_
 *  1  `Nelle Varoquaux`_
 *  1  Rafael Cunha de Almeida
 *  1  Rolando Espinoza La fuente
 *  1  `Vlad Niculae`_
 *  1  `Yaroslav Halchenko`_


.. _changes_0_13:

Version 0.13
============

**January 21, 2013**

New Estimator Classes
---------------------

- :class:`dummy.DummyClassifier` and :class:`dummy.DummyRegressor`, two
  data-independent predictors by `Mathieu Blondel`_. Useful to sanity-check
  your estimators. See :ref:`dummy_estimators` in the user guide.
  Multioutput support added by `Arnaud Joly`_.

- :class:`decomposition.FactorAnalysis`, a transformer implementing the
  classical factor analysis, by `Christian Osendorfer`_ and `Alexandre
  Gramfort`_. See :ref:`FA` in the user guide.

- :class:`feature_extraction.FeatureHasher`, a transformer implementing the
  "hashing trick" for fast, low-memory feature extraction from string fields
  by `Lars Buitinck`_ and :class:`feature_extraction.text.HashingVectorizer`
  for text documents by `Olivier Grisel`_  See :ref:`feature_hashing` and
  :ref:`hashing_vectorizer` for the documentation and sample usage.

- :class:`pipeline.FeatureUnion`, a transformer that concatenates
  results of several other transformers by `Andreas Müller`_. See
  :ref:`feature_union` in the user guide.

- :class:`random_projection.GaussianRandomProjection`,
  :class:`random_projection.SparseRandomProjection` and the function
  :func:`random_projection.johnson_lindenstrauss_min_dim`. The first two are
  transformers implementing Gaussian and sparse random projection matrix
  by `Olivier Grisel`_ and `Arnaud Joly`_.
  See :ref:`random_projection` in the user guide.

- :class:`kernel_approximation.Nystroem`, a transformer for approximating
  arbitrary kernels by `Andreas Müller`_. See
  :ref:`nystroem_kernel_approx` in the user guide.

- :class:`preprocessing.OneHotEncoder`, a transformer that computes binary
  encodings of categorical features by `Andreas Müller`_. See
  :ref:`preprocessing_categorical_features` in the user guide.

- :class:`linear_model.PassiveAggressiveClassifier` and
  :class:`linear_model.PassiveAggressiveRegressor`, predictors implementing
  an efficient stochastic optimization for linear models by `Rob Zinkov`_ and
  `Mathieu Blondel`_. See :ref:`passive_aggressive` in the user
  guide.

- :class:`ensemble.RandomTreesEmbedding`, a transformer for creating high-dimensional
  sparse representations using ensembles of totally random trees by  `Andreas Müller`_.
  See :ref:`random_trees_embedding` in the user guide.

- :class:`manifold.SpectralEmbedding` and function
  :func:`manifold.spectral_embedding`, implementing the "laplacian
  eigenmaps" transformation for non-linear dimensionality reduction by Wei
  Li. See :ref:`spectral_embedding` in the user guide.

- :class:`isotonic.IsotonicRegression` by `Fabian Pedregosa`_, `Alexandre Gramfort`_
  and `Nelle Varoquaux`_,


Changelog
---------

- :func:`metrics.zero_one_loss` (formerly ``metrics.zero_one``) now has
  option for normalized output that reports the fraction of
  misclassifications, rather than the raw number of misclassifications. By
  Kyle Beauchamp.

- :class:`tree.DecisionTreeClassifier` and all derived ensemble models now
  support sample weighting, by `Noel Dawe`_  and `Gilles Louppe`_.

- Speedup improvement when using bootstrap samples in forests of randomized
  trees, by `Peter Prettenhofer`_  and `Gilles Louppe`_.

- Partial dependence plots for :ref:`gradient_boosting` in
  :func:`ensemble.partial_dependence.partial_dependence` by `Peter
  Prettenhofer`_. See :ref:`sphx_glr_auto_examples_ensemble_plot_partial_dependence.py` for an
  example.

- The table of contents on the website has now been made expandable by
  `Jaques Grobler`_.

- :class:`feature_selection.SelectPercentile` now breaks ties
  deterministically instead of returning all equally ranked features.

- :class:`feature_selection.SelectKBest` and
  :class:`feature_selection.SelectPercentile` are more numerically stable
  since they use scores, rather than p-values, to rank results. This means
  that they might sometimes select different features than they did
  previously.

- Ridge regression and ridge classification fitting with ``sparse_cg`` solver
  no longer has quadratic memory complexity, by `Lars Buitinck`_ and
  `Fabian Pedregosa`_.

- Ridge regression and ridge classification now support a new fast solver
  called ``lsqr``, by `Mathieu Blondel`_.

- Speed up of :func:`metrics.precision_recall_curve` by Conrad Lee.

- Added support for reading/writing svmlight files with pairwise
  preference attribute (qid in svmlight file format) in
  :func:`datasets.dump_svmlight_file` and
  :func:`datasets.load_svmlight_file` by `Fabian Pedregosa`_.

- Faster and more robust :func:`metrics.confusion_matrix` and
  :ref:`clustering_evaluation` by Wei Li.

- :func:`cross_validation.cross_val_score` now works with precomputed kernels
  and affinity matrices, by `Andreas Müller`_.

- LARS algorithm made more numerically stable with heuristics to drop
  regressors too correlated as well as to stop the path when
  numerical noise becomes predominant, by `Gael Varoquaux`_.

- Faster implementation of :func:`metrics.precision_recall_curve` by
  Conrad Lee.

- New kernel :class:`metrics.chi2_kernel` by `Andreas Müller`_, often used
  in computer vision applications.

- Fix of longstanding bug in :class:`naive_bayes.BernoulliNB` fixed by
  Shaun Jackman.

- Implemented ``predict_proba`` in :class:`multiclass.OneVsRestClassifier`,
  by Andrew Winterman.

- Improve consistency in gradient boosting: estimators
  :class:`ensemble.GradientBoostingRegressor` and
  :class:`ensemble.GradientBoostingClassifier` use the estimator
  :class:`tree.DecisionTreeRegressor` instead of the
  :class:`tree._tree.Tree` data structure by `Arnaud Joly`_.

- Fixed a floating point exception in the :ref:`decision trees <tree>`
  module, by Seberg.

- Fix :func:`metrics.roc_curve` fails when y_true has only one class
  by Wei Li.

- Add the :func:`metrics.mean_absolute_error` function which computes the
  mean absolute error. The :func:`metrics.mean_squared_error`,
  :func:`metrics.mean_absolute_error` and
  :func:`metrics.r2_score` metrics support multioutput by `Arnaud Joly`_.

- Fixed ``class_weight`` support in :class:`svm.LinearSVC` and
  :class:`linear_model.LogisticRegression` by `Andreas Müller`_. The meaning
  of ``class_weight`` was reversed as erroneously higher weight meant less
  positives of a given class in earlier releases.

- Improve narrative documentation and consistency in
  :mod:`sklearn.metrics` for regression and classification metrics
  by `Arnaud Joly`_.

- Fixed a bug in :class:`sklearn.svm.SVC` when using csr-matrices with
  unsorted indices by Xinfan Meng and `Andreas Müller`_.

- :class:`MiniBatchKMeans`: Add random reassignment of cluster centers
  with little observations attached to them, by `Gael Varoquaux`_.


API changes summary
-------------------
- Renamed all occurrences of ``n_atoms`` to ``n_components`` for consistency.
  This applies to :class:`decomposition.DictionaryLearning`,
  :class:`decomposition.MiniBatchDictionaryLearning`,
  :func:`decomposition.dict_learning`, :func:`decomposition.dict_learning_online`.

- Renamed all occurrences of ``max_iters`` to ``max_iter`` for consistency.
  This applies to :class:`semi_supervised.LabelPropagation` and
  :class:`semi_supervised.label_propagation.LabelSpreading`.

- Renamed all occurrences of ``learn_rate`` to ``learning_rate`` for
  consistency in :class:`ensemble.BaseGradientBoosting` and
  :class:`ensemble.GradientBoostingRegressor`.

- The module ``sklearn.linear_model.sparse`` is gone. Sparse matrix support
  was already integrated into the "regular" linear models.

- :func:`sklearn.metrics.mean_square_error`, which incorrectly returned the
  accumulated error, was removed. Use ``mean_squared_error`` instead.

- Passing ``class_weight`` parameters to ``fit`` methods is no longer
  supported. Pass them to estimator constructors instead.

- GMMs no longer have ``decode`` and ``rvs`` methods. Use the ``score``,
  ``predict`` or ``sample`` methods instead.

- The ``solver`` fit option in Ridge regression and classification is now
  deprecated and will be removed in v0.14. Use the constructor option
  instead.

- :class:`feature_extraction.text.DictVectorizer` now returns sparse
  matrices in the CSR format, instead of COO.

- Renamed ``k`` in :class:`cross_validation.KFold` and
  :class:`cross_validation.StratifiedKFold` to ``n_folds``, renamed
  ``n_bootstraps`` to ``n_iter`` in ``cross_validation.Bootstrap``.

- Renamed all occurrences of ``n_iterations`` to ``n_iter`` for consistency.
  This applies to :class:`cross_validation.ShuffleSplit`,
  :class:`cross_validation.StratifiedShuffleSplit`,
  :func:`utils.randomized_range_finder` and :func:`utils.randomized_svd`.

- Replaced ``rho`` in :class:`linear_model.ElasticNet` and
  :class:`linear_model.SGDClassifier` by ``l1_ratio``. The ``rho`` parameter
  had different meanings; ``l1_ratio`` was introduced to avoid confusion.
  It has the same meaning as previously ``rho`` in
  :class:`linear_model.ElasticNet` and ``(1-rho)`` in
  :class:`linear_model.SGDClassifier`.

- :class:`linear_model.LassoLars` and :class:`linear_model.Lars` now
  store a list of paths in the case of multiple targets, rather than
  an array of paths.

- The attribute ``gmm`` of :class:`hmm.GMMHMM` was renamed to ``gmm_``
  to adhere more strictly with the API.

- :func:`cluster.spectral_embedding` was moved to
  :func:`manifold.spectral_embedding`.

- Renamed ``eig_tol`` in :func:`manifold.spectral_embedding`,
  :class:`cluster.SpectralClustering` to ``eigen_tol``, renamed ``mode``
  to ``eigen_solver``.

- Renamed ``mode`` in :func:`manifold.spectral_embedding` and
  :class:`cluster.SpectralClustering` to ``eigen_solver``.

- ``classes_`` and ``n_classes_`` attributes of
  :class:`tree.DecisionTreeClassifier` and all derived ensemble models are
  now flat in case of single output problems and nested in case of
  multi-output problems.

- The ``estimators_`` attribute of
  :class:`ensemble.gradient_boosting.GradientBoostingRegressor` and
  :class:`ensemble.gradient_boosting.GradientBoostingClassifier` is now an
  array of :class:'tree.DecisionTreeRegressor'.

- Renamed ``chunk_size`` to ``batch_size`` in
  :class:`decomposition.MiniBatchDictionaryLearning` and
  :class:`decomposition.MiniBatchSparsePCA` for consistency.

- :class:`svm.SVC` and :class:`svm.NuSVC` now provide a ``classes_``
  attribute and support arbitrary dtypes for labels ``y``.
  Also, the dtype returned by ``predict`` now reflects the dtype of
  ``y`` during ``fit`` (used to be ``np.float``).

- Changed default test_size in :func:`cross_validation.train_test_split`
  to None, added possibility to infer ``test_size`` from ``train_size`` in
  :class:`cross_validation.ShuffleSplit` and
  :class:`cross_validation.StratifiedShuffleSplit`.

- Renamed function :func:`sklearn.metrics.zero_one` to
  :func:`sklearn.metrics.zero_one_loss`. Be aware that the default behavior
  in :func:`sklearn.metrics.zero_one_loss` is different from
  :func:`sklearn.metrics.zero_one`: ``normalize=False`` is changed to
  ``normalize=True``.

- Renamed function :func:`metrics.zero_one_score` to
  :func:`metrics.accuracy_score`.

- :func:`datasets.make_circles` now has the same number of inner and outer points.

- In the Naive Bayes classifiers, the ``class_prior`` parameter was moved
  from ``fit`` to ``__init__``.

People
------
List of contributors for release 0.13 by number of commits.

 * 364  `Andreas Müller`_
 * 143  `Arnaud Joly`_
 * 137  `Peter Prettenhofer`_
 * 131  `Gael Varoquaux`_
 * 117  `Mathieu Blondel`_
 * 108  `Lars Buitinck`_
 * 106  Wei Li
 * 101  `Olivier Grisel`_
 *  65  `Vlad Niculae`_
 *  54  `Gilles Louppe`_
 *  40  `Jaques Grobler`_
 *  38  `Alexandre Gramfort`_
 *  30  `Rob Zinkov`_
 *  19  Aymeric Masurelle
 *  18  Andrew Winterman
 *  17  `Fabian Pedregosa`_
 *  17  Nelle Varoquaux
 *  16  `Christian Osendorfer`_
 *  14  `Daniel Nouri`_
 *  13  :user:`Virgile Fritsch <VirgileFritsch>`
 *  13  syhw
 *  12  `Satrajit Ghosh`_
 *  10  Corey Lynch
 *  10  Kyle Beauchamp
 *   9  Brian Cheung
 *   9  Immanuel Bayer
 *   9  mr.Shu
 *   8  Conrad Lee
 *   8  `James Bergstra`_
 *   7  Tadej Janež
 *   6  Brian Cajes
 *   6  `Jake Vanderplas`_
 *   6  Michael
 *   6  Noel Dawe
 *   6  Tiago Nunes
 *   6  cow
 *   5  Anze
 *   5  Shiqiao Du
 *   4  Christian Jauvin
 *   4  Jacques Kvam
 *   4  Richard T. Guy
 *   4  `Robert Layton`_
 *   3  Alexandre Abraham
 *   3  Doug Coleman
 *   3  Scott Dickerson
 *   2  ApproximateIdentity
 *   2  John Benediktsson
 *   2  Mark Veronda
 *   2  Matti Lyra
 *   2  Mikhail Korobov
 *   2  Xinfan Meng
 *   1  Alejandro Weinstein
 *   1  `Alexandre Passos`_
 *   1  Christoph Deil
 *   1  Eugene Nizhibitsky
 *   1  Kenneth C. Arnold
 *   1  Luis Pedro Coelho
 *   1  Miroslav Batchkarov
 *   1  Pavel
 *   1  Sebastian Berg
 *   1  Shaun Jackman
 *   1  Subhodeep Moitra
 *   1  bob
 *   1  dengemann
 *   1  emanuele
 *   1  x006


.. _changes_0_12.1:

Version 0.12.1
===============

**October 8, 2012**

The 0.12.1 release is a bug-fix release with no additional features, but is
instead a set of bug fixes

Changelog
----------

- Improved numerical stability in spectral embedding by `Gael
  Varoquaux`_

- Doctest under windows 64bit by `Gael Varoquaux`_

- Documentation fixes for elastic net by `Andreas Müller`_ and
  `Alexandre Gramfort`_

- Proper behavior with fortran-ordered NumPy arrays by `Gael Varoquaux`_

- Make GridSearchCV work with non-CSR sparse matrix by `Lars Buitinck`_

- Fix parallel computing in MDS by `Gael Varoquaux`_

- Fix Unicode support in count vectorizer by `Andreas Müller`_

- Fix MinCovDet breaking with X.shape = (3, 1) by :user:`Virgile Fritsch <VirgileFritsch>`

- Fix clone of SGD objects by `Peter Prettenhofer`_

- Stabilize GMM by :user:`Virgile Fritsch <VirgileFritsch>`

People
------

 *  14  `Peter Prettenhofer`_
 *  12  `Gael Varoquaux`_
 *  10  `Andreas Müller`_
 *   5  `Lars Buitinck`_
 *   3  :user:`Virgile Fritsch <VirgileFritsch>`
 *   1  `Alexandre Gramfort`_
 *   1  `Gilles Louppe`_
 *   1  `Mathieu Blondel`_

.. _changes_0_12:

Version 0.12
============

**September 4, 2012**

Changelog
---------

- Various speed improvements of the :ref:`decision trees <tree>` module, by
  `Gilles Louppe`_.

- :class:`ensemble.GradientBoostingRegressor` and
  :class:`ensemble.GradientBoostingClassifier` now support feature subsampling
  via the ``max_features`` argument, by `Peter Prettenhofer`_.

- Added Huber and Quantile loss functions to
  :class:`ensemble.GradientBoostingRegressor`, by `Peter Prettenhofer`_.

- :ref:`Decision trees <tree>` and :ref:`forests of randomized trees <forest>`
  now support multi-output classification and regression problems, by
  `Gilles Louppe`_.

- Added :class:`preprocessing.LabelEncoder`, a simple utility class to
  normalize labels or transform non-numerical labels, by `Mathieu Blondel`_.

- Added the epsilon-insensitive loss and the ability to make probabilistic
  predictions with the modified huber loss in :ref:`sgd`, by
  `Mathieu Blondel`_.

- Added :ref:`multidimensional_scaling`, by Nelle Varoquaux.

- SVMlight file format loader now detects compressed (gzip/bzip2) files and
  decompresses them on the fly, by `Lars Buitinck`_.

- SVMlight file format serializer now preserves double precision floating
  point values, by `Olivier Grisel`_.

- A common testing framework for all estimators was added, by `Andreas Müller`_.

- Understandable error messages for estimators that do not accept
  sparse input by `Gael Varoquaux`_

- Speedups in hierarchical clustering by `Gael Varoquaux`_. In
  particular building the tree now supports early stopping. This is
  useful when the number of clusters is not small compared to the
  number of samples.

- Add MultiTaskLasso and MultiTaskElasticNet for joint feature selection,
  by `Alexandre Gramfort`_.

- Added :func:`metrics.auc_score` and
  :func:`metrics.average_precision_score` convenience functions by `Andreas
  Müller`_.

- Improved sparse matrix support in the :ref:`feature_selection`
  module by `Andreas Müller`_.

- New word boundaries-aware character n-gram analyzer for the
  :ref:`text_feature_extraction` module by :user:`@kernc <kernc>`.

- Fixed bug in spectral clustering that led to single point clusters
  by `Andreas Müller`_.

- In :class:`feature_extraction.text.CountVectorizer`, added an option to
  ignore infrequent words, ``min_df`` by  `Andreas Müller`_.

- Add support for multiple targets in some linear models (ElasticNet, Lasso
  and OrthogonalMatchingPursuit) by `Vlad Niculae`_ and
  `Alexandre Gramfort`_.

- Fixes in :class:`decomposition.ProbabilisticPCA` score function by Wei Li.

- Fixed feature importance computation in
  :ref:`gradient_boosting`.

API changes summary
-------------------

- The old ``scikits.learn`` package has disappeared; all code should import
  from ``sklearn`` instead, which was introduced in 0.9.

- In :func:`metrics.roc_curve`, the ``thresholds`` array is now returned
  with it's order reversed, in order to keep it consistent with the order
  of the returned ``fpr`` and ``tpr``.

- In :class:`hmm` objects, like :class:`hmm.GaussianHMM`,
  :class:`hmm.MultinomialHMM`, etc., all parameters must be passed to the
  object when initialising it and not through ``fit``. Now ``fit`` will
  only accept the data as an input parameter.

- For all SVM classes, a faulty behavior of ``gamma`` was fixed. Previously,
  the default gamma value was only computed the first time ``fit`` was called
  and then stored. It is now recalculated on every call to ``fit``.

- All ``Base`` classes are now abstract meta classes so that they can not be
  instantiated.

- :func:`cluster.ward_tree` now also returns the parent array. This is
  necessary for early-stopping in which case the tree is not
  completely built.

- In :class:`feature_extraction.text.CountVectorizer` the parameters
  ``min_n`` and ``max_n`` were joined to the parameter ``n_gram_range`` to
  enable grid-searching both at once.

- In :class:`feature_extraction.text.CountVectorizer`, words that appear
  only in one document are now ignored by default. To reproduce
  the previous behavior, set ``min_df=1``.

- Fixed API inconsistency: :meth:`linear_model.SGDClassifier.predict_proba` now
  returns 2d array when fit on two classes.

- Fixed API inconsistency: :meth:`discriminant_analysis.QuadraticDiscriminantAnalysis.decision_function`
  and :meth:`discriminant_analysis.LinearDiscriminantAnalysis.decision_function` now return 1d arrays
  when fit on two classes.

- Grid of alphas used for fitting :class:`linear_model.LassoCV` and
  :class:`linear_model.ElasticNetCV` is now stored
  in the attribute ``alphas_`` rather than overriding the init parameter
  ``alphas``.

- Linear models when alpha is estimated by cross-validation store
  the estimated value in the ``alpha_`` attribute rather than just
  ``alpha`` or ``best_alpha``.

- :class:`ensemble.GradientBoostingClassifier` now supports
  :meth:`ensemble.GradientBoostingClassifier.staged_predict_proba`, and
  :meth:`ensemble.GradientBoostingClassifier.staged_predict`.

- :class:`svm.sparse.SVC` and other sparse SVM classes are now deprecated.
  The all classes in the :ref:`svm` module now automatically select the
  sparse or dense representation base on the input.

- All clustering algorithms now interpret the array ``X`` given to ``fit`` as
  input data, in particular :class:`cluster.SpectralClustering` and
  :class:`cluster.AffinityPropagation` which previously expected affinity matrices.

- For clustering algorithms that take the desired number of clusters as a parameter,
  this parameter is now called ``n_clusters``.


People
------
 * 267  `Andreas Müller`_
 *  94  `Gilles Louppe`_
 *  89  `Gael Varoquaux`_
 *  79  `Peter Prettenhofer`_
 *  60  `Mathieu Blondel`_
 *  57  `Alexandre Gramfort`_
 *  52  `Vlad Niculae`_
 *  45  `Lars Buitinck`_
 *  44  Nelle Varoquaux
 *  37  `Jaques Grobler`_
 *  30  Alexis Mignon
 *  30  Immanuel Bayer
 *  27  `Olivier Grisel`_
 *  16  Subhodeep Moitra
 *  13  Yannick Schwartz
 *  12  :user:`@kernc <kernc>`
 *  11  :user:`Virgile Fritsch <VirgileFritsch>`
 *   9  Daniel Duckworth
 *   9  `Fabian Pedregosa`_
 *   9  `Robert Layton`_
 *   8  John Benediktsson
 *   7  Marko Burjek
 *   5  `Nicolas Pinto`_
 *   4  Alexandre Abraham
 *   4  `Jake Vanderplas`_
 *   3  `Brian Holt`_
 *   3  `Edouard Duchesnay`_
 *   3  Florian Hoenig
 *   3  flyingimmidev
 *   2  Francois Savard
 *   2  Hannes Schulz
 *   2  Peter Welinder
 *   2  `Yaroslav Halchenko`_
 *   2  Wei Li
 *   1  Alex Companioni
 *   1  Brandyn A. White
 *   1  Bussonnier Matthias
 *   1  Charles-Pierre Astolfi
 *   1  Dan O'Huiginn
 *   1  David Cournapeau
 *   1  Keith Goodman
 *   1  Ludwig Schwardt
 *   1  Olivier Hervieu
 *   1  Sergio Medina
 *   1  Shiqiao Du
 *   1  Tim Sheerman-Chase
 *   1  buguen



.. _changes_0_11:

Version 0.11
============

**May 7, 2012**

Changelog
---------

Highlights
.............

- Gradient boosted regression trees (:ref:`gradient_boosting`)
  for classification and regression by `Peter Prettenhofer`_
  and `Scott White`_ .

- Simple dict-based feature loader with support for categorical variables
  (:class:`feature_extraction.DictVectorizer`) by `Lars Buitinck`_.

- Added Matthews correlation coefficient (:func:`metrics.matthews_corrcoef`)
  and added macro and micro average options to
  :func:`metrics.precision_score`, :func:`metrics.recall_score` and
  :func:`metrics.f1_score` by `Satrajit Ghosh`_.

- :ref:`out_of_bag` of generalization error for :ref:`ensemble`
  by `Andreas Müller`_.

- Randomized sparse linear models for feature
  selection, by `Alexandre Gramfort`_ and `Gael Varoquaux`_

- :ref:`label_propagation` for semi-supervised learning, by Clay
  Woolam. **Note** the semi-supervised API is still work in progress,
  and may change.

- Added BIC/AIC model selection to classical :ref:`gmm` and unified
  the API with the remainder of scikit-learn, by `Bertrand Thirion`_

- Added :class:`sklearn.cross_validation.StratifiedShuffleSplit`, which is
  a :class:`sklearn.cross_validation.ShuffleSplit` with balanced splits,
  by Yannick Schwartz.

- :class:`sklearn.neighbors.NearestCentroid` classifier added, along with a
  ``shrink_threshold`` parameter, which implements **shrunken centroid
  classification**, by `Robert Layton`_.

Other changes
..............

- Merged dense and sparse implementations of :ref:`sgd` module and
  exposed utility extension types for sequential
  datasets ``seq_dataset`` and weight vectors ``weight_vector``
  by `Peter Prettenhofer`_.

- Added ``partial_fit`` (support for online/minibatch learning) and
  warm_start to the :ref:`sgd` module by `Mathieu Blondel`_.

- Dense and sparse implementations of :ref:`svm` classes and
  :class:`linear_model.LogisticRegression` merged by `Lars Buitinck`_.

- Regressors can now be used as base estimator in the :ref:`multiclass`
  module by `Mathieu Blondel`_.

- Added n_jobs option to :func:`metrics.pairwise.pairwise_distances`
  and :func:`metrics.pairwise.pairwise_kernels` for parallel computation,
  by `Mathieu Blondel`_.

- :ref:`k_means` can now be run in parallel, using the ``n_jobs`` argument
  to either :ref:`k_means` or :class:`KMeans`, by `Robert Layton`_.

- Improved :ref:`cross_validation` and :ref:`grid_search` documentation
  and introduced the new :func:`cross_validation.train_test_split`
  helper function by `Olivier Grisel`_

- :class:`svm.SVC` members ``coef_`` and ``intercept_`` changed sign for
  consistency with ``decision_function``; for ``kernel==linear``,
  ``coef_`` was fixed in the one-vs-one case, by `Andreas Müller`_.

- Performance improvements to efficient leave-one-out cross-validated
  Ridge regression, esp. for the ``n_samples > n_features`` case, in
  :class:`linear_model.RidgeCV`, by Reuben Fletcher-Costin.

- Refactoring and simplification of the :ref:`text_feature_extraction`
  API and fixed a bug that caused possible negative IDF,
  by `Olivier Grisel`_.

- Beam pruning option in :class:`_BaseHMM` module has been removed since it
  is difficult to Cythonize. If you are interested in contributing a Cython
  version, you can use the python version in the git history as a reference.

- Classes in :ref:`neighbors` now support arbitrary Minkowski metric for
  nearest neighbors searches. The metric can be specified by argument ``p``.

API changes summary
-------------------

- :class:`covariance.EllipticEnvelop` is now deprecated - Please use :class:`covariance.EllipticEnvelope`
  instead.

- ``NeighborsClassifier`` and ``NeighborsRegressor`` are gone in the module
  :ref:`neighbors`. Use the classes :class:`KNeighborsClassifier`,
  :class:`RadiusNeighborsClassifier`, :class:`KNeighborsRegressor`
  and/or :class:`RadiusNeighborsRegressor` instead.

- Sparse classes in the :ref:`sgd` module are now deprecated.

- In :class:`mixture.GMM`, :class:`mixture.DPGMM` and :class:`mixture.VBGMM`,
  parameters must be passed to an object when initialising it and not through
  ``fit``. Now ``fit`` will only accept the data as an input parameter.

- methods ``rvs`` and ``decode`` in :class:`GMM` module are now deprecated.
  ``sample`` and ``score`` or ``predict`` should be used instead.

- attribute ``_scores`` and ``_pvalues`` in univariate feature selection
  objects are now deprecated.
  ``scores_`` or ``pvalues_`` should be used instead.

- In :class:`LogisticRegression`, :class:`LinearSVC`, :class:`SVC` and
  :class:`NuSVC`, the ``class_weight`` parameter is now an initialization
  parameter, not a parameter to fit. This makes grid searches
  over this parameter possible.

- LFW ``data`` is now always shape ``(n_samples, n_features)`` to be
  consistent with the Olivetti faces dataset. Use ``images`` and
  ``pairs`` attribute to access the natural images shapes instead.

- In :class:`svm.LinearSVC`, the meaning of the ``multi_class`` parameter
  changed.  Options now are ``'ovr'`` and ``'crammer_singer'``, with
  ``'ovr'`` being the default.  This does not change the default behavior
  but hopefully is less confusing.

- Class :class:`feature_selection.text.Vectorizer` is deprecated and
  replaced by :class:`feature_selection.text.TfidfVectorizer`.

- The preprocessor / analyzer nested structure for text feature
  extraction has been removed. All those features are
  now directly passed as flat constructor arguments
  to :class:`feature_selection.text.TfidfVectorizer` and
  :class:`feature_selection.text.CountVectorizer`, in particular the
  following parameters are now used:

- ``analyzer`` can be ``'word'`` or ``'char'`` to switch the default
  analysis scheme, or use a specific python callable (as previously).

- ``tokenizer`` and ``preprocessor`` have been introduced to make it
  still possible to customize those steps with the new API.

- ``input`` explicitly control how to interpret the sequence passed to
  ``fit`` and ``predict``: filenames, file objects or direct (byte or
  Unicode) strings.

- charset decoding is explicit and strict by default.

- the ``vocabulary``, fitted or not is now stored in the
  ``vocabulary_`` attribute to be consistent with the project
  conventions.

- Class :class:`feature_selection.text.TfidfVectorizer` now derives directly
  from :class:`feature_selection.text.CountVectorizer` to make grid
  search trivial.

- methods ``rvs`` in :class:`_BaseHMM` module are now deprecated.
  ``sample`` should be used instead.

- Beam pruning option in :class:`_BaseHMM` module is removed since it is
  difficult to be Cythonized. If you are interested, you can look in the
  history codes by git.

- The SVMlight format loader now supports files with both zero-based and
  one-based column indices, since both occur "in the wild".

- Arguments in class :class:`ShuffleSplit` are now consistent with
  :class:`StratifiedShuffleSplit`. Arguments ``test_fraction`` and
  ``train_fraction`` are deprecated and renamed to ``test_size`` and
  ``train_size`` and can accept both ``float`` and ``int``.

- Arguments in class :class:`Bootstrap` are now consistent with
  :class:`StratifiedShuffleSplit`. Arguments ``n_test`` and
  ``n_train`` are deprecated and renamed to ``test_size`` and
  ``train_size`` and can accept both ``float`` and ``int``.

- Argument ``p`` added to classes in :ref:`neighbors` to specify an
  arbitrary Minkowski metric for nearest neighbors searches.


People
------
   * 282  `Andreas Müller`_
   * 239  `Peter Prettenhofer`_
   * 198  `Gael Varoquaux`_
   * 129  `Olivier Grisel`_
   * 114  `Mathieu Blondel`_
   * 103  Clay Woolam
   *  96  `Lars Buitinck`_
   *  88  `Jaques Grobler`_
   *  82  `Alexandre Gramfort`_
   *  50  `Bertrand Thirion`_
   *  42  `Robert Layton`_
   *  28  flyingimmidev
   *  26  `Jake Vanderplas`_
   *  26  Shiqiao Du
   *  21  `Satrajit Ghosh`_
   *  17  `David Marek`_
   *  17  `Gilles Louppe`_
   *  14  `Vlad Niculae`_
   *  11  Yannick Schwartz
   *  10  `Fabian Pedregosa`_
   *   9  fcostin
   *   7  Nick Wilson
   *   5  Adrien Gaidon
   *   5  `Nicolas Pinto`_
   *   4  `David Warde-Farley`_
   *   5  Nelle Varoquaux
   *   5  Emmanuelle Gouillart
   *   3  Joonas Sillanpää
   *   3  Paolo Losi
   *   2  Charles McCarthy
   *   2  Roy Hyunjin Han
   *   2  Scott White
   *   2  ibayer
   *   1  Brandyn White
   *   1  Carlos Scheidegger
   *   1  Claire Revillet
   *   1  Conrad Lee
   *   1  `Edouard Duchesnay`_
   *   1  Jan Hendrik Metzen
   *   1  Meng Xinfan
   *   1  `Rob Zinkov`_
   *   1  Shiqiao
   *   1  Udi Weinsberg
   *   1  Virgile Fritsch
   *   1  Xinfan Meng
   *   1  Yaroslav Halchenko
   *   1  jansoe
   *   1  Leon Palafox


.. _changes_0_10:

Version 0.10
============

**January 11, 2012**

Changelog
---------

- Python 2.5 compatibility was dropped; the minimum Python version needed
  to use scikit-learn is now 2.6.

- :ref:`sparse_inverse_covariance` estimation using the graph Lasso, with
  associated cross-validated estimator, by `Gael Varoquaux`_

- New :ref:`Tree <tree>` module by `Brian Holt`_, `Peter Prettenhofer`_,
  `Satrajit Ghosh`_ and `Gilles Louppe`_. The module comes with complete
  documentation and examples.

- Fixed a bug in the RFE module by `Gilles Louppe`_ (issue #378).

- Fixed a memory leak in :ref:`svm` module by `Brian Holt`_ (issue #367).

- Faster tests by `Fabian Pedregosa`_ and others.

- Silhouette Coefficient cluster analysis evaluation metric added as
  :func:`sklearn.metrics.silhouette_score` by Robert Layton.

- Fixed a bug in :ref:`k_means` in the handling of the ``n_init`` parameter:
  the clustering algorithm used to be run ``n_init`` times but the last
  solution was retained instead of the best solution by `Olivier Grisel`_.

- Minor refactoring in :ref:`sgd` module; consolidated dense and sparse
  predict methods; Enhanced test time performance by converting model
  parameters to fortran-style arrays after fitting (only multi-class).

- Adjusted Mutual Information metric added as
  :func:`sklearn.metrics.adjusted_mutual_info_score` by Robert Layton.

- Models like SVC/SVR/LinearSVC/LogisticRegression from libsvm/liblinear
  now support scaling of C regularization parameter by the number of
  samples by `Alexandre Gramfort`_.

- New :ref:`Ensemble Methods <ensemble>` module by `Gilles Louppe`_ and
  `Brian Holt`_. The module comes with the random forest algorithm and the
  extra-trees method, along with documentation and examples.

- :ref:`outlier_detection`: outlier and novelty detection, by
  :user:`Virgile Fritsch <VirgileFritsch>`.

- :ref:`kernel_approximation`: a transform implementing kernel
  approximation for fast SGD on non-linear kernels by
  `Andreas Müller`_.

- Fixed a bug due to atom swapping in :ref:`OMP` by `Vlad Niculae`_.

- :ref:`SparseCoder` by `Vlad Niculae`_.

- :ref:`mini_batch_kmeans` performance improvements by `Olivier Grisel`_.

- :ref:`k_means` support for sparse matrices by `Mathieu Blondel`_.

- Improved documentation for developers and for the :mod:`sklearn.utils`
  module, by `Jake Vanderplas`_.

- Vectorized 20newsgroups dataset loader
  (:func:`sklearn.datasets.fetch_20newsgroups_vectorized`) by
  `Mathieu Blondel`_.

- :ref:`multiclass` by `Lars Buitinck`_.

- Utilities for fast computation of mean and variance for sparse matrices
  by `Mathieu Blondel`_.

- Make :func:`sklearn.preprocessing.scale` and
  :class:`sklearn.preprocessing.Scaler` work on sparse matrices by
  `Olivier Grisel`_

- Feature importances using decision trees and/or forest of trees,
  by `Gilles Louppe`_.

- Parallel implementation of forests of randomized trees by
  `Gilles Louppe`_.

- :class:`sklearn.cross_validation.ShuffleSplit` can subsample the train
  sets as well as the test sets by `Olivier Grisel`_.

- Errors in the build of the documentation fixed by `Andreas Müller`_.


API changes summary
-------------------

Here are the code migration instructions when upgrading from scikit-learn
version 0.9:

- Some estimators that may overwrite their inputs to save memory previously
  had ``overwrite_`` parameters; these have been replaced with ``copy_``
  parameters with exactly the opposite meaning.

  This particularly affects some of the estimators in :mod:`linear_model`.
  The default behavior is still to copy everything passed in.

- The SVMlight dataset loader :func:`sklearn.datasets.load_svmlight_file` no
  longer supports loading two files at once; use ``load_svmlight_files``
  instead. Also, the (unused) ``buffer_mb`` parameter is gone.

- Sparse estimators in the :ref:`sgd` module use dense parameter vector
  ``coef_`` instead of ``sparse_coef_``. This significantly improves
  test time performance.

- The :ref:`covariance` module now has a robust estimator of
  covariance, the Minimum Covariance Determinant estimator.

- Cluster evaluation metrics in :mod:`metrics.cluster` have been refactored
  but the changes are backwards compatible. They have been moved to the
  :mod:`metrics.cluster.supervised`, along with
  :mod:`metrics.cluster.unsupervised` which contains the Silhouette
  Coefficient.

- The ``permutation_test_score`` function now behaves the same way as
  ``cross_val_score`` (i.e. uses the mean score across the folds.)

- Cross Validation generators now use integer indices (``indices=True``)
  by default instead of boolean masks. This make it more intuitive to
  use with sparse matrix data.

- The functions used for sparse coding, ``sparse_encode`` and
  ``sparse_encode_parallel`` have been combined into
  :func:`sklearn.decomposition.sparse_encode`, and the shapes of the arrays
  have been transposed for consistency with the matrix factorization setting,
  as opposed to the regression setting.

- Fixed an off-by-one error in the SVMlight/LibSVM file format handling;
  files generated using :func:`sklearn.datasets.dump_svmlight_file` should be
  re-generated. (They should continue to work, but accidentally had one
  extra column of zeros prepended.)

- ``BaseDictionaryLearning`` class replaced by ``SparseCodingMixin``.

- :func:`sklearn.utils.extmath.fast_svd` has been renamed
  :func:`sklearn.utils.extmath.randomized_svd` and the default
  oversampling is now fixed to 10 additional random vectors instead
  of doubling the number of components to extract. The new behavior
  follows the reference paper.


People
------

The following people contributed to scikit-learn since last release:

   * 246  `Andreas Müller`_
   * 242  `Olivier Grisel`_
   * 220  `Gilles Louppe`_
   * 183  `Brian Holt`_
   * 166  `Gael Varoquaux`_
   * 144  `Lars Buitinck`_
   *  73  `Vlad Niculae`_
   *  65  `Peter Prettenhofer`_
   *  64  `Fabian Pedregosa`_
   *  60  Robert Layton
   *  55  `Mathieu Blondel`_
   *  52  `Jake Vanderplas`_
   *  44  Noel Dawe
   *  38  `Alexandre Gramfort`_
   *  24  :user:`Virgile Fritsch <VirgileFritsch>`
   *  23  `Satrajit Ghosh`_
   *   3  Jan Hendrik Metzen
   *   3  Kenneth C. Arnold
   *   3  Shiqiao Du
   *   3  Tim Sheerman-Chase
   *   3  `Yaroslav Halchenko`_
   *   2  Bala Subrahmanyam Varanasi
   *   2  DraXus
   *   2  Michael Eickenberg
   *   1  Bogdan Trach
   *   1  Félix-Antoine Fortin
   *   1  Juan Manuel Caicedo Carvajal
   *   1  Nelle Varoquaux
   *   1  `Nicolas Pinto`_
   *   1  Tiziano Zito
   *   1  Xinfan Meng



.. _changes_0_9:

Version 0.9
===========

**September 21, 2011**

scikit-learn 0.9 was released on September 2011, three months after the 0.8
release and includes the new modules :ref:`manifold`, :ref:`dirichlet_process`
as well as several new algorithms and documentation improvements.

This release also includes the dictionary-learning work developed by
`Vlad Niculae`_ as part of the `Google Summer of Code
<https://developers.google.com/open-source/gsoc>`_ program.



.. |banner1| image:: ./auto_examples/manifold/images/thumb/sphx_glr_plot_compare_methods_thumb.png
   :target: auto_examples/manifold/plot_compare_methods.html

.. |banner2| image:: ./auto_examples/linear_model/images/thumb/sphx_glr_plot_omp_thumb.png
   :target: auto_examples/linear_model/plot_omp.html

.. |banner3| image:: ./auto_examples/decomposition/images/thumb/sphx_glr_plot_kernel_pca_thumb.png
   :target: auto_examples/decomposition/plot_kernel_pca.html

.. |center-div| raw:: html

    <div style="text-align: center; margin: 0px 0 -5px 0;">

.. |end-div| raw:: html

    </div>


|center-div| |banner2| |banner1| |banner3| |end-div|

Changelog
---------

- New :ref:`manifold` module by `Jake Vanderplas`_ and
  `Fabian Pedregosa`_.

- New :ref:`Dirichlet Process <dirichlet_process>` Gaussian Mixture
  Model by `Alexandre Passos`_

- :ref:`neighbors` module refactoring by `Jake Vanderplas`_ :
  general refactoring, support for sparse matrices in input, speed and
  documentation improvements. See the next section for a full list of API
  changes.

- Improvements on the :ref:`feature_selection` module by
  `Gilles Louppe`_ : refactoring of the RFE classes, documentation
  rewrite, increased efficiency and minor API changes.

- :ref:`SparsePCA` by `Vlad Niculae`_, `Gael Varoquaux`_ and
  `Alexandre Gramfort`_

- Printing an estimator now behaves independently of architectures
  and Python version thanks to :user:`Jean Kossaifi <JeanKossaifi>`.

- :ref:`Loader for libsvm/svmlight format <libsvm_loader>` by
  `Mathieu Blondel`_ and `Lars Buitinck`_

- Documentation improvements: thumbnails in
  example gallery by `Fabian Pedregosa`_.

- Important bugfixes in :ref:`svm` module (segfaults, bad
  performance) by `Fabian Pedregosa`_.

- Added :ref:`multinomial_naive_bayes` and :ref:`bernoulli_naive_bayes`
  by `Lars Buitinck`_

- Text feature extraction optimizations by Lars Buitinck

- Chi-Square feature selection
  (:func:`feature_selection.univariate_selection.chi2`) by `Lars Buitinck`_.

- :ref:`sample_generators` module refactoring by `Gilles Louppe`_

- :ref:`multiclass` by `Mathieu Blondel`_

- Ball tree rewrite by `Jake Vanderplas`_

- Implementation of :ref:`dbscan` algorithm by Robert Layton

- Kmeans predict and transform by Robert Layton

- Preprocessing module refactoring by `Olivier Grisel`_

- Faster mean shift by Conrad Lee

- New ``Bootstrap``, :ref:`ShuffleSplit` and various other
  improvements in cross validation schemes by `Olivier Grisel`_ and
  `Gael Varoquaux`_

- Adjusted Rand index and V-Measure clustering evaluation metrics by `Olivier Grisel`_

- Added :class:`Orthogonal Matching Pursuit <linear_model.OrthogonalMatchingPursuit>` by `Vlad Niculae`_

- Added 2D-patch extractor utilities in the :ref:`feature_extraction` module by `Vlad Niculae`_

- Implementation of :class:`linear_model.LassoLarsCV`
  (cross-validated Lasso solver using the Lars algorithm) and
  :class:`linear_model.LassoLarsIC` (BIC/AIC model
  selection in Lars) by `Gael Varoquaux`_
  and `Alexandre Gramfort`_

- Scalability improvements to :func:`metrics.roc_curve` by Olivier Hervieu

- Distance helper functions :func:`metrics.pairwise.pairwise_distances`
  and :func:`metrics.pairwise.pairwise_kernels` by Robert Layton

- :class:`Mini-Batch K-Means <cluster.MiniBatchKMeans>` by Nelle Varoquaux and Peter Prettenhofer.

- :ref:`mldata` utilities by Pietro Berkes.

- :ref:`olivetti_faces` by `David Warde-Farley`_.


API changes summary
-------------------

Here are the code migration instructions when upgrading from scikit-learn
version 0.8:

- The ``scikits.learn`` package was renamed ``sklearn``. There is
  still a ``scikits.learn`` package alias for backward compatibility.

  Third-party projects with a dependency on scikit-learn 0.9+ should
  upgrade their codebase. For instance, under Linux / MacOSX just run
  (make a backup first!)::

      find -name "*.py" | xargs sed -i 's/\bscikits.learn\b/sklearn/g'

- Estimators no longer accept model parameters as ``fit`` arguments:
  instead all parameters must be only be passed as constructor
  arguments or using the now public ``set_params`` method inherited
  from :class:`base.BaseEstimator`.

  Some estimators can still accept keyword arguments on the ``fit``
  but this is restricted to data-dependent values (e.g. a Gram matrix
  or an affinity matrix that are precomputed from the ``X`` data matrix.

- The ``cross_val`` package has been renamed to ``cross_validation``
  although there is also a ``cross_val`` package alias in place for
  backward compatibility.

  Third-party projects with a dependency on scikit-learn 0.9+ should
  upgrade their codebase. For instance, under Linux / MacOSX just run
  (make a backup first!)::

      find -name "*.py" | xargs sed -i 's/\bcross_val\b/cross_validation/g'

- The ``score_func`` argument of the
  ``sklearn.cross_validation.cross_val_score`` function is now expected
  to accept ``y_test`` and ``y_predicted`` as only arguments for
  classification and regression tasks or ``X_test`` for unsupervised
  estimators.

- ``gamma`` parameter for support vector machine algorithms is set
  to ``1 / n_features`` by default, instead of ``1 / n_samples``.

- The ``sklearn.hmm`` has been marked as orphaned: it will be removed
  from scikit-learn in version 0.11 unless someone steps up to
  contribute documentation, examples and fix lurking numerical
  stability issues.

- ``sklearn.neighbors`` has been made into a submodule.  The two previously
  available estimators, ``NeighborsClassifier`` and ``NeighborsRegressor``
  have been marked as deprecated.  Their functionality has been divided
  among five new classes: ``NearestNeighbors`` for unsupervised neighbors
  searches, ``KNeighborsClassifier`` & ``RadiusNeighborsClassifier``
  for supervised classification problems, and ``KNeighborsRegressor``
  & ``RadiusNeighborsRegressor`` for supervised regression problems.

- ``sklearn.ball_tree.BallTree`` has been moved to
  ``sklearn.neighbors.BallTree``.  Using the former will generate a warning.

- ``sklearn.linear_model.LARS()`` and related classes (LassoLARS,
  LassoLARSCV, etc.) have been renamed to
  ``sklearn.linear_model.Lars()``.

- All distance metrics and kernels in ``sklearn.metrics.pairwise`` now have a Y
  parameter, which by default is None. If not given, the result is the distance
  (or kernel similarity) between each sample in Y. If given, the result is the
  pairwise distance (or kernel similarity) between samples in X to Y.

- ``sklearn.metrics.pairwise.l1_distance`` is now called ``manhattan_distance``,
  and by default returns the pairwise distance. For the component wise distance,
  set the parameter ``sum_over_features`` to ``False``.

Backward compatibility package aliases and other deprecated classes and
functions will be removed in version 0.11.


People
------

38 people contributed to this release.

- 387  `Vlad Niculae`_
- 320  `Olivier Grisel`_
- 192  `Lars Buitinck`_
- 179  `Gael Varoquaux`_
- 168  `Fabian Pedregosa`_ (`INRIA`_, `Parietal Team`_)
- 127  `Jake Vanderplas`_
- 120  `Mathieu Blondel`_
- 85  `Alexandre Passos`_
- 67  `Alexandre Gramfort`_
- 57  `Peter Prettenhofer`_
- 56  `Gilles Louppe`_
- 42  Robert Layton
- 38  Nelle Varoquaux
- 32  :user:`Jean Kossaifi <JeanKossaifi>`
- 30  Conrad Lee
- 22  Pietro Berkes
- 18  andy
- 17  David Warde-Farley
- 12  Brian Holt
- 11  Robert
- 8  Amit Aides
- 8  :user:`Virgile Fritsch <VirgileFritsch>`
- 7  `Yaroslav Halchenko`_
- 6  Salvatore Masecchia
- 5  Paolo Losi
- 4  Vincent Schut
- 3  Alexis Metaireau
- 3  Bryan Silverthorn
- 3  `Andreas Müller`_
- 2  Minwoo Jake Lee
- 1  Emmanuelle Gouillart
- 1  Keith Goodman
- 1  Lucas Wiman
- 1  `Nicolas Pinto`_
- 1  Thouis (Ray) Jones
- 1  Tim Sheerman-Chase


.. _changes_0_8:

Version 0.8
===========

**May 11, 2011**

scikit-learn 0.8 was released on May 2011, one month after the first
"international" `scikit-learn coding sprint
<https://github.com/scikit-learn/scikit-learn/wiki/Upcoming-events>`_ and is
marked by the inclusion of important modules: :ref:`hierarchical_clustering`,
:ref:`cross_decomposition`, :ref:`NMF`, initial support for Python 3 and by important
enhancements and bug fixes.


Changelog
---------

Several new modules where introduced during this release:

- New :ref:`hierarchical_clustering` module by Vincent Michel,
  `Bertrand Thirion`_, `Alexandre Gramfort`_ and `Gael Varoquaux`_.

- :ref:`kernel_pca` implementation by `Mathieu Blondel`_

- :ref:`labeled_faces_in_the_wild` by `Olivier Grisel`_.

- New :ref:`cross_decomposition` module by `Edouard Duchesnay`_.

- :ref:`NMF` module `Vlad Niculae`_

- Implementation of the :ref:`oracle_approximating_shrinkage` algorithm by
  :user:`Virgile Fritsch <VirgileFritsch>` in the :ref:`covariance` module.


Some other modules benefited from significant improvements or cleanups.


- Initial support for Python 3: builds and imports cleanly,
  some modules are usable while others have failing tests by `Fabian Pedregosa`_.

- :class:`decomposition.PCA` is now usable from the Pipeline object by `Olivier Grisel`_.

- Guide :ref:`performance-howto` by `Olivier Grisel`_.

- Fixes for memory leaks in libsvm bindings, 64-bit safer BallTree by Lars Buitinck.

- bug and style fixing in :ref:`k_means` algorithm by Jan Schlüter.

- Add attribute converged to Gaussian Mixture Models by Vincent Schut.

- Implemented ``transform``, ``predict_log_proba`` in
  :class:`discriminant_analysis.LinearDiscriminantAnalysis` By `Mathieu Blondel`_.

- Refactoring in the :ref:`svm` module and bug fixes by `Fabian Pedregosa`_,
  `Gael Varoquaux`_ and Amit Aides.

- Refactored SGD module (removed code duplication, better variable naming),
  added interface for sample weight by `Peter Prettenhofer`_.

- Wrapped BallTree with Cython by Thouis (Ray) Jones.

- Added function :func:`svm.l1_min_c` by Paolo Losi.

- Typos, doc style, etc. by `Yaroslav Halchenko`_, `Gael Varoquaux`_,
  `Olivier Grisel`_, Yann Malet, `Nicolas Pinto`_, Lars Buitinck and
  `Fabian Pedregosa`_.


People
-------

People that made this release possible preceded by number of commits:


- 159  `Olivier Grisel`_
- 96  `Gael Varoquaux`_
- 96  `Vlad Niculae`_
- 94  `Fabian Pedregosa`_
- 36  `Alexandre Gramfort`_
- 32  Paolo Losi
- 31  `Edouard Duchesnay`_
- 30  `Mathieu Blondel`_
- 25  `Peter Prettenhofer`_
- 22  `Nicolas Pinto`_
- 11  :user:`Virgile Fritsch <VirgileFritsch>`
   -  7  Lars Buitinck
   -  6  Vincent Michel
   -  5  `Bertrand Thirion`_
   -  4  Thouis (Ray) Jones
   -  4  Vincent Schut
   -  3  Jan Schlüter
   -  2  Julien Miotte
   -  2  `Matthieu Perrot`_
   -  2  Yann Malet
   -  2  `Yaroslav Halchenko`_
   -  1  Amit Aides
   -  1  `Andreas Müller`_
   -  1  Feth Arezki
   -  1  Meng Xinfan


.. _changes_0_7:

Version 0.7
===========

**March 2, 2011**

scikit-learn 0.7 was released in March 2011, roughly three months
after the 0.6 release. This release is marked by the speed
improvements in existing algorithms like k-Nearest Neighbors and
K-Means algorithm and by the inclusion of an efficient algorithm for
computing the Ridge Generalized Cross Validation solution. Unlike the
preceding release, no new modules where added to this release.

Changelog
---------

- Performance improvements for Gaussian Mixture Model sampling [Jan
  Schlüter].

- Implementation of efficient leave-one-out cross-validated Ridge in
  :class:`linear_model.RidgeCV` [`Mathieu Blondel`_]

- Better handling of collinearity and early stopping in
  :func:`linear_model.lars_path` [`Alexandre Gramfort`_ and `Fabian
  Pedregosa`_].

- Fixes for liblinear ordering of labels and sign of coefficients
  [Dan Yamins, Paolo Losi, `Mathieu Blondel`_ and `Fabian Pedregosa`_].

- Performance improvements for Nearest Neighbors algorithm in
  high-dimensional spaces [`Fabian Pedregosa`_].

- Performance improvements for :class:`cluster.KMeans` [`Gael
  Varoquaux`_ and `James Bergstra`_].

- Sanity checks for SVM-based classes [`Mathieu Blondel`_].

- Refactoring of :class:`neighbors.NeighborsClassifier` and
  :func:`neighbors.kneighbors_graph`: added different algorithms for
  the k-Nearest Neighbor Search and implemented a more stable
  algorithm for finding barycenter weights. Also added some
  developer documentation for this module, see
  `notes_neighbors
  <https://github.com/scikit-learn/scikit-learn/wiki/Neighbors-working-notes>`_ for more information [`Fabian Pedregosa`_].

- Documentation improvements: Added :class:`pca.RandomizedPCA` and
  :class:`linear_model.LogisticRegression` to the class
  reference. Also added references of matrices used for clustering
  and other fixes [`Gael Varoquaux`_, `Fabian Pedregosa`_, `Mathieu
  Blondel`_, `Olivier Grisel`_, Virgile Fritsch , Emmanuelle
  Gouillart]

- Binded decision_function in classes that make use of liblinear_,
  dense and sparse variants, like :class:`svm.LinearSVC` or
  :class:`linear_model.LogisticRegression` [`Fabian Pedregosa`_].

- Performance and API improvements to
  :func:`metrics.euclidean_distances` and to
  :class:`pca.RandomizedPCA` [`James Bergstra`_].

- Fix compilation issues under NetBSD [Kamel Ibn Hassen Derouiche]

- Allow input sequences of different lengths in :class:`hmm.GaussianHMM`
  [`Ron Weiss`_].

- Fix bug in affinity propagation caused by incorrect indexing [Xinfan Meng]


People
------

People that made this release possible preceded by number of commits:

- 85  `Fabian Pedregosa`_
- 67  `Mathieu Blondel`_
- 20  `Alexandre Gramfort`_
- 19  `James Bergstra`_
- 14  Dan Yamins
- 13  `Olivier Grisel`_
- 12  `Gael Varoquaux`_
- 4  `Edouard Duchesnay`_
- 4  `Ron Weiss`_
- 2  Satrajit Ghosh
- 2  Vincent Dubourg
- 1  Emmanuelle Gouillart
- 1  Kamel Ibn Hassen Derouiche
- 1  Paolo Losi
- 1  VirgileFritsch
- 1  `Yaroslav Halchenko`_
- 1  Xinfan Meng


.. _changes_0_6:

Version 0.6
===========

**December 21, 2010**

scikit-learn 0.6 was released on December 2010. It is marked by the
inclusion of several new modules and a general renaming of old
ones. It is also marked by the inclusion of new example, including
applications to real-world datasets.


Changelog
---------

- New `stochastic gradient
  <http://scikit-learn.org/stable/modules/sgd.html>`_ descent
  module by Peter Prettenhofer. The module comes with complete
  documentation and examples.

- Improved svm module: memory consumption has been reduced by 50%,
  heuristic to automatically set class weights, possibility to
  assign weights to samples (see
  :ref:`sphx_glr_auto_examples_svm_plot_weighted_samples.py` for an example).

- New :ref:`gaussian_process` module by Vincent Dubourg. This module
  also has great documentation and some very neat examples. See
  example_gaussian_process_plot_gp_regression.py or
  example_gaussian_process_plot_gp_probabilistic_classification_after_regression.py
  for a taste of what can be done.

- It is now possible to use liblinear’s Multi-class SVC (option
  multi_class in :class:`svm.LinearSVC`)

- New features and performance improvements of text feature
  extraction.

- Improved sparse matrix support, both in main classes
  (:class:`grid_search.GridSearchCV`) as in modules
  sklearn.svm.sparse and sklearn.linear_model.sparse.

- Lots of cool new examples and a new section that uses real-world
  datasets was created. These include:
  :ref:`sphx_glr_auto_examples_applications_plot_face_recognition.py`,
  :ref:`sphx_glr_auto_examples_applications_plot_species_distribution_modeling.py`,
  :ref:`sphx_glr_auto_examples_applications_svm_gui.py`,
  :ref:`sphx_glr_auto_examples_applications_wikipedia_principal_eigenvector.py` and
  others.

- Faster :ref:`least_angle_regression` algorithm. It is now 2x
  faster than the R version on worst case and up to 10x times faster
  on some cases.

- Faster coordinate descent algorithm. In particular, the full path
  version of lasso (:func:`linear_model.lasso_path`) is more than
  200x times faster than before.

- It is now possible to get probability estimates from a
  :class:`linear_model.LogisticRegression` model.

- module renaming: the glm module has been renamed to linear_model,
  the gmm module has been included into the more general mixture
  model and the sgd module has been included in linear_model.

- Lots of bug fixes and documentation improvements.


People
------

People that made this release possible preceded by number of commits:

   * 207  `Olivier Grisel`_

   * 167 `Fabian Pedregosa`_

   * 97 `Peter Prettenhofer`_

   * 68 `Alexandre Gramfort`_

   * 59  `Mathieu Blondel`_

   * 55  `Gael Varoquaux`_

   * 33  Vincent Dubourg

   * 21  `Ron Weiss`_

   * 9  Bertrand Thirion

   * 3  `Alexandre Passos`_

   * 3  Anne-Laure Fouque

   * 2  Ronan Amicel

   * 1 `Christian Osendorfer`_



.. _changes_0_5:


Version 0.5
===========

**October 11, 2010**

Changelog
---------

New classes
-----------

- Support for sparse matrices in some classifiers of modules
  ``svm`` and ``linear_model`` (see :class:`svm.sparse.SVC`,
  :class:`svm.sparse.SVR`, :class:`svm.sparse.LinearSVC`,
  :class:`linear_model.sparse.Lasso`, :class:`linear_model.sparse.ElasticNet`)

- New :class:`pipeline.Pipeline` object to compose different estimators.

- Recursive Feature Elimination routines in module
  :ref:`feature_selection`.

- Addition of various classes capable of cross validation in the
  linear_model module (:class:`linear_model.LassoCV`, :class:`linear_model.ElasticNetCV`,
  etc.).

- New, more efficient LARS algorithm implementation. The Lasso
  variant of the algorithm is also implemented. See
  :class:`linear_model.lars_path`, :class:`linear_model.Lars` and
  :class:`linear_model.LassoLars`.

- New Hidden Markov Models module (see classes
  :class:`hmm.GaussianHMM`, :class:`hmm.MultinomialHMM`,
  :class:`hmm.GMMHMM`)

- New module feature_extraction (see :ref:`class reference
  <feature_extraction_ref>`)

- New FastICA algorithm in module sklearn.fastica


Documentation
-------------

- Improved documentation for many modules, now separating
  narrative documentation from the class reference. As an example,
  see `documentation for the SVM module
  <http://scikit-learn.org/stable/modules/svm.html>`_ and the
  complete `class reference
  <http://scikit-learn.org/stable/modules/classes.html>`_.

Fixes
-----

- API changes: adhere variable names to PEP-8, give more
  meaningful names.

- Fixes for svm module to run on a shared memory context
  (multiprocessing).

- It is again possible to generate latex (and thus PDF) from the
  sphinx docs.

Examples
--------

- new examples using some of the mlcomp datasets:
  ``sphx_glr_auto_examples_mlcomp_sparse_document_classification.py`` (since removed) and
  :ref:`sphx_glr_auto_examples_text_document_classification_20newsgroups.py`

- Many more examples. `See here
  <http://scikit-learn.org/stable/auto_examples/index.html>`_
  the full list of examples.


External dependencies
---------------------

- Joblib is now a dependency of this package, although it is
  shipped with (sklearn.externals.joblib).

Removed modules
---------------

- Module ann (Artificial Neural Networks) has been removed from
  the distribution. Users wanting this sort of algorithms should
  take a look into pybrain.

Misc
----

- New sphinx theme for the web page.


Authors
-------

The following is a list of authors for this release, preceded by
number of commits:

     * 262  Fabian Pedregosa
     * 240  Gael Varoquaux
     * 149  Alexandre Gramfort
     * 116  Olivier Grisel
     *  40  Vincent Michel
     *  38  Ron Weiss
     *  23  Matthieu Perrot
     *  10  Bertrand Thirion
     *   7  Yaroslav Halchenko
     *   9  VirgileFritsch
     *   6  Edouard Duchesnay
     *   4  Mathieu Blondel
     *   1  Ariel Rokem
     *   1  Matthieu Brucher

Version 0.4
===========

**August 26, 2010**

Changelog
---------

Major changes in this release include:

- Coordinate Descent algorithm (Lasso, ElasticNet) refactoring &
  speed improvements (roughly 100x times faster).

- Coordinate Descent Refactoring (and bug fixing) for consistency
  with R's package GLMNET.

- New metrics module.

- New GMM module contributed by Ron Weiss.

- Implementation of the LARS algorithm (without Lasso variant for now).

- feature_selection module redesign.

- Migration to GIT as version control system.

- Removal of obsolete attrselect module.

- Rename of private compiled extensions (added underscore).

- Removal of legacy unmaintained code.

- Documentation improvements (both docstring and rst).

- Improvement of the build system to (optionally) link with MKL.
  Also, provide a lite BLAS implementation in case no system-wide BLAS is
  found.

- Lots of new examples.

- Many, many bug fixes ...


Authors
-------

The committer list for this release is the following (preceded by number
of commits):

    * 143  Fabian Pedregosa
    * 35  Alexandre Gramfort
    * 34  Olivier Grisel
    * 11  Gael Varoquaux
    *  5  Yaroslav Halchenko
    *  2  Vincent Michel
    *  1  Chris Filo Gorgolewski


Earlier versions
================

Earlier versions included contributions by Fred Mailhot, David Cooke,
David Huard, Dave Morrill, Ed Schofield, Travis Oliphant, Pearu Peterson.

.. _Olivier Grisel: https://twitter.com/ogrisel

.. _Gael Varoquaux: http://gael-varoquaux.info

.. _Alexandre Gramfort: http://alexandre.gramfort.net

.. _Fabian Pedregosa: http://fa.bianp.net

.. _Mathieu Blondel: http://www.mblondel.org

.. _James Bergstra: http://www-etud.iro.umontreal.ca/~bergstrj/

.. _liblinear: http://www.csie.ntu.edu.tw/~cjlin/liblinear/

.. _Yaroslav Halchenko: http://www.onerussian.com/

.. _Vlad Niculae: http://vene.ro

.. _Edouard Duchesnay: https://sites.google.com/site/duchesnay/home

.. _Peter Prettenhofer: https://sites.google.com/site/peterprettenhofer/

.. _Alexandre Passos: http://atpassos.me

.. _Nicolas Pinto: https://twitter.com/npinto

.. _Bertrand Thirion: https://team.inria.fr/parietal/bertrand-thirions-page

.. _Andreas Müller: http://peekaboo-vision.blogspot.com

.. _Matthieu Perrot: http://brainvisa.info/biblio/lnao/en/Author/PERROT-M.html

.. _Jake Vanderplas: http://staff.washington.edu/jakevdp/

.. _Gilles Louppe: http://www.montefiore.ulg.ac.be/~glouppe/

.. _INRIA: http://www.inria.fr

.. _Parietal Team: http://parietal.saclay.inria.fr/

.. _David Warde-Farley: http://www-etud.iro.umontreal.ca/~wardefar/

.. _Brian Holt: http://personal.ee.surrey.ac.uk/Personal/B.Holt

.. _Satrajit Ghosh: http://www.mit.edu/~satra/

.. _Robert Layton: https://twitter.com/robertlayton

.. _Scott White: https://twitter.com/scottblanc

.. _David Marek: http://www.davidmarek.cz/

.. _Christian Osendorfer: https://osdf.github.io

.. _Arnaud Joly: http://www.ajoly.org

.. _Rob Zinkov: http://zinkov.com

.. _Joel Nothman: http://joelnothman.com

.. _Nicolas Trésegnie : http://nicolastr.com/

.. _Kemal Eren: http://www.kemaleren.com

.. _Yann Dauphin: http://ynd.github.io/

.. _Yannick Schwartz: https://team.inria.fr/parietal/schwarty/

.. _Kyle Kastner: http://kastnerkyle.github.io

.. _Daniel Nouri: http://danielnouri.org

.. _Manoj Kumar: https://manojbits.wordpress.com

.. _Luis Pedro Coelho: http://luispedro.org

.. _Fares Hedyati: http://www.eecs.berkeley.edu/~fareshed

.. _Antony Lee: https://www.ocf.berkeley.edu/~antonyl/

.. _Martin Billinger: http://tnsre.embs.org/author/martinbillinger

.. _Matteo Visconti di Oleggio Castello: http://www.mvdoc.me

.. _Trevor Stephens: http://trevorstephens.com/

.. _Jan Hendrik Metzen: https://jmetzen.github.io/

.. _Will Dawson: http://www.dawsonresearch.com

.. _Andrew Tulloch: http://tullo.ch/

.. _Hanna Wallach: http://dirichlet.net/

.. _Yan Yi: http://seowyanyi.org

.. _Hervé Bredin: http://herve.niderb.fr/

.. _Eric Martin: http://www.ericmart.in

.. _Nicolas Goix: https://perso.telecom-paristech.fr/~goix/

.. _Sebastian Raschka: http://sebastianraschka.com

.. _Brian McFee: https://bmcfee.github.io

.. _Valentin Stolbunov: http://www.vstolbunov.com

.. _Jaques Grobler: https://github.com/jaquesgrobler

.. _Lars Buitinck: https://github.com/larsmans

.. _Loic Esteve: https://github.com/lesteve

.. _Noel Dawe: https://github.com/ndawe

.. _Raghav RV: https://github.com/raghavrv

.. _Tom Dupre la Tour: https://github.com/TomDLT

.. _Nelle Varoquaux: https://github.com/nellev

.. _Bing Tian Dai: https://github.com/btdai

.. _Dylan Werner-Meier: https://github.com/unautre

.. _Alyssa Batula: https://github.com/abatula

.. _Srivatsan Ramesh: https://github.com/srivatsan-ramesh

.. _Ron Weiss: http://www.ee.columbia.edu/~ronw

.. _Kathleen Chen: https://github.com/kchen17

.. _Vincent Pham: https://github.com/vincentpham1991

.. _Denis Engemann: http://denis-engemann.de
.. _Anish Shah: https://github.com/AnishShah

.. _Neeraj Gangwar: http://neerajgangwar.in
.. _Arthur Mensch: https://amensch.fr
