<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
<title>0.12 — scikit-learn 0.12 documentation</title>
<link rel="stylesheet" href="_static/nature.css" type="text/css" />
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<script type="text/javascript">
var DOCUMENTATION_OPTIONS = {
URL_ROOT: '',
VERSION: '0.12',
COLLAPSE_INDEX: false,
FILE_SUFFIX: '.html',
HAS_SOURCE: true
};
</script>
<script type="text/javascript" src="_static/jquery.js"></script>
<script type="text/javascript" src="_static/underscore.js"></script>
<script type="text/javascript" src="_static/doctools.js"></script>
<script type="text/javascript" src="_static/sidebar.js"></script>
<link rel="shortcut icon" href="_static/favicon.ico"/>
<link rel="author" title="About these documents" href="about.html" />
<link rel="top" title="scikit-learn 0.12 documentation" href="index.html" />
<link rel="next" title="Presentations and Tutorials on Scikit-Learn" href="presentations.html" />
<link rel="prev" title="Support" href="support.html" />
<script type="text/javascript">
var _gaq = _gaq || [];
_gaq.push(['_setAccount', 'UA-22606712-2']);
_gaq.push(['_trackPageview']);
(function() {
var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true;
ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js';
var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s);
})();
</script>
</head>
<body>
<div class="header-wrapper">
<div class="header">
<p class="logo"><a href="index.html">
<img src="_static/scikit-learn-logo-small.png" alt="Logo"/>
</a>
</p><div class="navbar">
<ul>
<li><a href="install.html">Download</a></li>
<li><a href="support.html">Support</a></li>
<li><a href="user_guide.html">User Guide</a></li>
<li><a href="auto_examples/index.html">Examples</a></li>
<li><a href="modules/classes.html">Reference</a></li>
</ul>
<div class="search_form">
<div id="cse" style="width: 100%;"></div>
<script src="http://www.google.com/jsapi" type="text/javascript"></script>
<script type="text/javascript">
google.load('search', '1', {language : 'en'});
google.setOnLoadCallback(function() {
var customSearchControl = new google.search.CustomSearchControl('016639176250731907682:tjtqbvtvij0');
customSearchControl.setResultSetSize(google.search.Search.FILTERED_CSE_RESULTSET);
var options = new google.search.DrawOptions();
options.setAutoComplete(true);
customSearchControl.draw('cse', options);
}, true);
</script>
</div>
</div> <!-- end navbar --></div>
</div>
<div class="content-wrapper">
<div class="sphinxsidebar">
<div class="sphinxsidebarwrapper">
<div class="rel rellarge">
<!-- rellinks[1:] is an ugly hack to avoid link to module
index -->
<div class="rellink">
<a href="support.html" title="Support"
accesskey="P">Previous
<br>
<span class="smallrellink">
Support
</span>
<span class="hiddenrellink">
Support
</span>
</a>
</div>
<div class="spacer">
</div>
<div class="rellink">
<a href="presentations.html" title="Presentations and Tutorials on Scikit-Learn"
accesskey="N">Next
<br>
<span class="smallrellink">
Presentations an...
</span>
<span class="hiddenrellink">
Presentations and Tutorials on Scikit-Learn
</span>
</a>
</div>
<!-- Ad a link to the 'up' page -->
</div>
<p style="text-align: center; background-color: #FFE4E4">This documentation is
for scikit-learn <strong>version 0.12</strong>
— <a href="http://scikit-learn.org/stable/support.html#documentation-resources">Other versions</a></p>
<h3>Citing</h3>
<p>If you use the software, please consider
<a href="about.html#citing-scikit-learn">citing scikit-learn</a>.</p>
<h3>This page</h3>
<ul>
<li><a class="reference internal" href="#">0.12</a><ul>
<li><a class="reference internal" href="#changelog">Changelog</a></li>
<li><a class="reference internal" href="#api-changes-summary">API changes summary</a></li>
<li><a class="reference internal" href="#people">People</a></li>
</ul>
</li>
<li><a class="reference internal" href="#changes-0-11">0.11</a><ul>
<li><a class="reference internal" href="#id3">Changelog</a><ul>
<li><a class="reference internal" href="#highlights">Highlights</a></li>
<li><a class="reference internal" href="#other-changes">Other changes</a></li>
</ul>
</li>
<li><a class="reference internal" href="#id4">API changes summary</a></li>
<li><a class="reference internal" href="#id5">People</a></li>
</ul>
</li>
<li><a class="reference internal" href="#changes-0-10">0.10</a><ul>
<li><a class="reference internal" href="#id7">Changelog</a></li>
<li><a class="reference internal" href="#id8">API changes summary</a></li>
<li><a class="reference internal" href="#id9">People</a></li>
</ul>
</li>
<li><a class="reference internal" href="#changes-0-9">0.9</a><ul>
<li><a class="reference internal" href="#id11">Changelog</a></li>
<li><a class="reference internal" href="#id12">API changes summary</a></li>
<li><a class="reference internal" href="#id13">People</a></li>
</ul>
</li>
<li><a class="reference internal" href="#changes-0-8">0.8</a><ul>
<li><a class="reference internal" href="#id15">Changelog</a></li>
<li><a class="reference internal" href="#id16">People</a></li>
</ul>
</li>
<li><a class="reference internal" href="#changes-0-7">0.7</a><ul>
<li><a class="reference internal" href="#id18">Changelog</a></li>
<li><a class="reference internal" href="#id19">People</a></li>
</ul>
</li>
<li><a class="reference internal" href="#changes-0-6">0.6</a><ul>
<li><a class="reference internal" href="#id21">Changelog</a></li>
<li><a class="reference internal" href="#id22">People</a></li>
</ul>
</li>
<li><a class="reference internal" href="#changes-0-5">0.5</a><ul>
<li><a class="reference internal" href="#id24">Changelog</a></li>
<li><a class="reference internal" href="#new-classes">New classes</a></li>
<li><a class="reference internal" href="#documentation">Documentation</a></li>
<li><a class="reference internal" href="#fixes">Fixes</a></li>
<li><a class="reference internal" href="#examples">Examples</a></li>
<li><a class="reference internal" href="#external-dependencies">External dependencies</a></li>
<li><a class="reference internal" href="#removed-modules">Removed modules</a></li>
<li><a class="reference internal" href="#misc">Misc</a></li>
<li><a class="reference internal" href="#authors">Authors</a></li>
</ul>
</li>
<li><a class="reference internal" href="#id25">0.4</a><ul>
<li><a class="reference internal" href="#id26">Changelog</a></li>
<li><a class="reference internal" href="#id27">Authors</a></li>
</ul>
</li>
</ul>
</div>
</div>
<div class="content">
<div class="documentwrapper">
<div class="bodywrapper">
<div class="body">
<div class="section" id="changes-0-12">
<span id="id1"></span><h1>0.12<a class="headerlink" href="#changes-0-12" title="Permalink to this headline">¶</a></h1>
<div class="section" id="changelog">
<h2>Changelog<a class="headerlink" href="#changelog" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>Various speed improvements of the <a class="reference internal" href="modules/tree.html#tree"><em>decision trees</em></a> module, by
<a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a>.</li>
<li><a class="reference internal" href="modules/generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor" title="sklearn.ensemble.GradientBoostingRegressor"><tt class="xref py py-class docutils literal"><span class="pre">ensemble.GradientBoostingRegressor</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><tt class="xref py py-class docutils literal"><span class="pre">ensemble.GradientBoostingClassifier</span></tt></a> now support feature subsampling
via the <tt class="docutils literal"><span class="pre">max_features</span></tt> argument, by <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a>.</li>
<li>Added Huber and Quantile loss functions to
<a class="reference internal" href="modules/generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor" title="sklearn.ensemble.GradientBoostingRegressor"><tt class="xref py py-class docutils literal"><span class="pre">ensemble.GradientBoostingRegressor</span></tt></a>, by <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a>.</li>
<li><a class="reference internal" href="modules/tree.html#tree"><em>Decision trees</em></a> and <a class="reference internal" href="modules/ensemble.html#forest"><em>forests of randomized trees</em></a>
now support multi-output classification and regression problems, by
<a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a>.</li>
<li>Added <a class="reference internal" href="modules/generated/sklearn.preprocessing.LabelEncoder.html#sklearn.preprocessing.LabelEncoder" title="sklearn.preprocessing.LabelEncoder"><tt class="xref py py-class docutils literal"><span class="pre">preprocessing.LabelEncoder</span></tt></a>, a simple utility class to
normalize labels or transform non-numerical labels, by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li>Added the epsilon-insensitive loss and the ability to make probabilistic
predictions with the modified huber loss in <a class="reference internal" href="modules/sgd.html#sgd"><em>Stochastic Gradient Descent</em></a>, by
<a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li>Added <a class="reference internal" href="modules/manifold.html#multidimensional-scaling"><em>Multi-dimensional Scaling (MDS)</em></a>, by Nelle Varoquaux.</li>
<li>SVMlight file format loader now detects compressed (gzip/bzip2) files and
decompresses them on the fly, by <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a>.</li>
<li>SVMlight file format serializer now preserves double precision floating
point values, by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li>A common testing framework for all estimators was added, by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
<li>Understandable error messages for estimators that do not accept
sparse input by <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>Speedups in hierarchical clustering by <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a>. 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.</li>
<li>Add MultiTaskLasso and MultiTaskElasticNet for joint feature selection,
by <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a>.</li>
<li>Added <a class="reference internal" href="modules/generated/sklearn.metrics.auc_score.html#sklearn.metrics.auc_score" title="sklearn.metrics.auc_score"><tt class="xref py py-func docutils literal"><span class="pre">metrics.auc_score</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.metrics.average_precision_score.html#sklearn.metrics.average_precision_score" title="sklearn.metrics.average_precision_score"><tt class="xref py py-func docutils literal"><span class="pre">metrics.average_precision_score</span></tt></a> convenience functions by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas
Müller</a>.</li>
<li>Improved sparse matrix support in the <a class="reference internal" href="modules/feature_selection.html#feature-selection"><em>Feature selection</em></a>
module by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
<li>New word boundaries-aware character n-gram analyzer for the
<a class="reference internal" href="modules/feature_extraction.html#text-feature-extraction"><em>Text feature extraction</em></a> module by <a class="reference external" href="http://github.com/kernc">@kernc</a>.</li>
<li>Fixed bug in spectral clustering that led to single point clusters
by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
<li>In <a class="reference internal" href="modules/generated/sklearn.feature_extraction.text.CountVectorizer.html#sklearn.feature_extraction.text.CountVectorizer" title="sklearn.feature_extraction.text.CountVectorizer"><tt class="xref py py-class docutils literal"><span class="pre">feature_extraction.text.CountVectorizer</span></tt></a>, added an option to
infrequent words, <tt class="docutils literal"><span class="pre">min_df</span></tt> by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
<li>Add support for multiple targets in some linear models (ElasticNet, Lasso
and OrthogonalMatchingPursuit) by <a class="reference external" href="http://vene.ro">Vlad Niculae</a> and
<a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a>.</li>
<li>Fixes in <a class="reference internal" href="modules/generated/sklearn.decomposition.ProbabilisticPCA.html#sklearn.decomposition.ProbabilisticPCA" title="sklearn.decomposition.ProbabilisticPCA"><tt class="xref py py-class docutils literal"><span class="pre">decomposition.ProbabilisticPCA</span></tt></a> score function by Wei Li.</li>
<li>Fixed feature importance computation in
<a class="reference internal" href="modules/ensemble.html#gradient-boosting"><em>Gradient Tree Boosting</em></a>.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="api-changes-summary">
<h2>API changes summary<a class="headerlink" href="#api-changes-summary" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>The old <tt class="docutils literal"><span class="pre">scikits.learn</span></tt> package has disappeared; all code should import
from <tt class="docutils literal"><span class="pre">sklearn</span></tt> instead, which was introduced in 0.9.</li>
<li>In <a class="reference internal" href="modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve" title="sklearn.metrics.roc_curve"><tt class="xref py py-func docutils literal"><span class="pre">metrics.roc_curve</span></tt></a>, the <tt class="docutils literal"><span class="pre">thresholds</span></tt> array is now returned
with it’s order reversed, in order to keep it consistent with the order
of the returned <tt class="docutils literal"><span class="pre">fpr</span></tt> and <tt class="docutils literal"><span class="pre">tpr</span></tt>.</li>
<li>In <a class="reference internal" href="modules/classes.html#module-sklearn.hmm" title="sklearn.hmm"><tt class="xref py py-class docutils literal"><span class="pre">hmm</span></tt></a> objects, like <a class="reference internal" href="modules/generated/sklearn.hmm.GaussianHMM.html#sklearn.hmm.GaussianHMM" title="sklearn.hmm.GaussianHMM"><tt class="xref py py-class docutils literal"><span class="pre">hmm.GaussianHMM</span></tt></a>,
<a class="reference internal" href="modules/generated/sklearn.hmm.MultinomialHMM.html#sklearn.hmm.MultinomialHMM" title="sklearn.hmm.MultinomialHMM"><tt class="xref py py-class docutils literal"><span class="pre">hmm.MultinomialHMM</span></tt></a>, etc., all parameters must be passed to the
object when initialising it and not through <tt class="docutils literal"><span class="pre">fit</span></tt>. Now <tt class="docutils literal"><span class="pre">fit</span></tt> will
only accept the data as an input parameter.</li>
<li>For all SVM classes, a faulty behavior of <tt class="docutils literal"><span class="pre">gamma</span></tt> was fixed. Previously,
the default gamma value was only computed the first time <tt class="docutils literal"><span class="pre">fit</span></tt> was called
and then stored. It is now recalculated on every call to <tt class="docutils literal"><span class="pre">fit</span></tt>.</li>
<li>All <tt class="docutils literal"><span class="pre">Base</span></tt> classes are now abstract meta classes so that they can not be
instantiated.</li>
<li><a class="reference internal" href="modules/generated/sklearn.cluster.ward_tree.html#sklearn.cluster.ward_tree" title="sklearn.cluster.ward_tree"><tt class="xref py py-func docutils literal"><span class="pre">cluster.ward_tree</span></tt></a> now also returns the parent array. This is
necessary for early-stopping in which case the tree is not
completely built.</li>
<li>In <a class="reference internal" href="modules/generated/sklearn.feature_extraction.text.CountVectorizer.html#sklearn.feature_extraction.text.CountVectorizer" title="sklearn.feature_extraction.text.CountVectorizer"><tt class="xref py py-class docutils literal"><span class="pre">feature_extraction.text.CountVectorizer</span></tt></a> the parameters
<tt class="docutils literal"><span class="pre">min_n</span></tt> and <tt class="docutils literal"><span class="pre">max_n</span></tt> were joined to the parameter <tt class="docutils literal"><span class="pre">n_gram_range</span></tt> to
enable grid-searching both at once.</li>
<li>In <a class="reference internal" href="modules/generated/sklearn.feature_extraction.text.CountVectorizer.html#sklearn.feature_extraction.text.CountVectorizer" title="sklearn.feature_extraction.text.CountVectorizer"><tt class="xref py py-class docutils literal"><span class="pre">feature_extraction.text.CountVectorizer</span></tt></a>, words that appear
only in one document are now ignored by default. To reproduce
the previous behavior, set <tt class="docutils literal"><span class="pre">min_df=1</span></tt>.</li>
<li>Fixed API inconsistency: <a class="reference internal" href="modules/generated/sklearn.linear_model.SGDClassifier.html#sklearn.linear_model.SGDClassifier.predict_proba" title="sklearn.linear_model.SGDClassifier.predict_proba"><tt class="xref py py-meth docutils literal"><span class="pre">linear_model.SGDClassifier.predict_proba</span></tt></a> now
returns 2d array when fit on two classes.</li>
<li>Fixed API inconsistency: <a class="reference internal" href="modules/generated/sklearn.qda.QDA.html#sklearn.qda.QDA.decision_function" title="sklearn.qda.QDA.decision_function"><tt class="xref py py-meth docutils literal"><span class="pre">qda.QDA.decision_function</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.lda.LDA.html#sklearn.lda.LDA.decision_function" title="sklearn.lda.LDA.decision_function"><tt class="xref py py-meth docutils literal"><span class="pre">lda.LDA.decision_function</span></tt></a> now return 1d arrays when fit on two
classes.</li>
<li>Grid of alphas used for fitting <a class="reference internal" href="modules/generated/sklearn.linear_model.LassoCV.html#sklearn.linear_model.LassoCV" title="sklearn.linear_model.LassoCV"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LassoCV</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.linear_model.ElasticNetCV.html#sklearn.linear_model.ElasticNetCV" title="sklearn.linear_model.ElasticNetCV"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.ElasticNetCV</span></tt></a> is now stored
in the attribute <cite>alphas_</cite> rather than overriding the init parameter
<cite>alphas</cite>.</li>
<li>Linear models when alpha is estimated by cross-validation store
the estimated value in the <cite>alpha_</cite> attribute rather than just
<cite>alpha</cite> or <cite>best_alpha</cite>.</li>
<li><a class="reference internal" href="modules/generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><tt class="xref py py-class docutils literal"><span class="pre">ensemble.GradientBoostingClassifier</span></tt></a> now supports
<a class="reference internal" href="modules/generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier.staged_predict_proba" title="sklearn.ensemble.GradientBoostingClassifier.staged_predict_proba"><tt class="xref py py-meth docutils literal"><span class="pre">ensemble.GradientBoostingClassifier.staged_predict_proba</span></tt></a>, and
<a class="reference internal" href="modules/generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier.staged_predict" title="sklearn.ensemble.GradientBoostingClassifier.staged_predict"><tt class="xref py py-meth docutils literal"><span class="pre">ensemble.GradientBoostingClassifier.staged_predict</span></tt></a>.</li>
<li><tt class="xref py py-class docutils literal"><span class="pre">svm.sparse.SVC</span></tt> and other sparse SVM classes are now deprecated.
The all classes in the <a class="reference internal" href="modules/svm.html#svm"><em>Support Vector Machines</em></a> module now automatically select the
sparse or dense representation base on the input.</li>
<li>All clustering algorithms now interpret the array <tt class="docutils literal"><span class="pre">X</span></tt> given to <tt class="docutils literal"><span class="pre">fit</span></tt> as
input data, in particular <a class="reference internal" href="modules/generated/sklearn.cluster.SpectralClustering.html#sklearn.cluster.SpectralClustering" title="sklearn.cluster.SpectralClustering"><tt class="xref py py-class docutils literal"><span class="pre">cluster.SpectralClustering</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.cluster.AffinityPropagation.html#sklearn.cluster.AffinityPropagation" title="sklearn.cluster.AffinityPropagation"><tt class="xref py py-class docutils literal"><span class="pre">cluster.AffinityPropagation</span></tt></a> which previously expected affinity matrices.</li>
<li>For clustering algorithms that take the desired number of clusters as a parameter,
this parameter is now called <tt class="docutils literal"><span class="pre">n_clusters</span></tt>.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="people">
<h2>People<a class="headerlink" href="#people" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>267 <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a></li>
<li>94 <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a></li>
<li>89 <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>79 <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a></li>
<li>60 <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>57 <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>52 <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>45 <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a></li>
<li>44 Nelle Varoquaux</li>
<li>37 <a class="reference external" href="https://github.com/jaquesgrobler/scikit-learn/wiki/Jaques-Grobler">Jaques Grobler</a></li>
<li>30 Alexis Mignon</li>
<li>30 Immanuel Bayer</li>
<li>27 <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>16 Subhodeep Moitra</li>
<li>13 Yannick Schwartz</li>
<li>12 <a class="reference external" href="http://github.com/kernc">@kernc</a></li>
<li>11 <a class="reference external" href="http://parietal.saclay.inria.fr/Members/virgile-fritsch">Virgile Fritsch</a></li>
<li>9 Daniel Duckworth</li>
<li>9 <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a></li>
<li>9 <a class="reference external" href="http://www.twitter.com/robertlayton">Robert Layton</a></li>
<li>8 John Benediktsson</li>
<li>7 Marko Burjek</li>
<li>5 <a class="reference external" href="http://pinto.scripts.mit.edu/">Nicolas Pinto</a></li>
<li>4 Alexandre Abraham</li>
<li>4 Jake VanderPlas</li>
<li>3 <a class="reference external" href="http://info.ee.surrey.ac.uk/Personal/B.Holt/">Brian Holt</a></li>
<li>3 <a class="reference external" href="http://www.lnao.fr/spip.php?rubrique30">Edouard Duchesnay</a></li>
<li>3 Florian Hoenig</li>
<li>3 flyingimmidev</li>
<li>2 Francois Savard</li>
<li>2 Hannes Schulz</li>
<li>2 Peter Welinder</li>
<li>2 <a class="reference external" href="http://www.onerussian.com/">Yaroslav Halchenko</a></li>
<li>2 Wei Li</li>
<li>1 Alex Companioni</li>
<li>1 Brandyn A. White</li>
<li>1 Bussonnier Matthias</li>
<li>1 Charles-Pierre Astolfi</li>
<li>1 Dan O’Huiginn</li>
<li>1 David Cournapeau</li>
<li>1 Keith Goodman</li>
<li>1 Ludwig Schwardt</li>
<li>1 Olivier Hervieu</li>
<li>1 Sergio Medina</li>
<li>1 Shiqiao Du</li>
<li>1 Tim Sheerman-Chase</li>
<li>1 buguen</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="changes-0-11">
<span id="id2"></span><h1>0.11<a class="headerlink" href="#changes-0-11" title="Permalink to this headline">¶</a></h1>
<div class="section" id="id3">
<h2>Changelog<a class="headerlink" href="#id3" title="Permalink to this headline">¶</a></h2>
<div class="section" id="highlights">
<h3>Highlights<a class="headerlink" href="#highlights" title="Permalink to this headline">¶</a></h3>
<blockquote>
<div><ul class="simple">
<li>Gradient boosted regression trees (<a class="reference internal" href="modules/ensemble.html#gradient-boosting"><em>Gradient Tree Boosting</em></a>)
for classification and regression by <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a>
and <a class="reference external" href="http://twitter.com/scottblanc">Scott White</a> .</li>
<li>Simple dict-based feature loader with support for categorical variables
(<a class="reference internal" href="modules/generated/sklearn.feature_extraction.DictVectorizer.html#sklearn.feature_extraction.DictVectorizer" title="sklearn.feature_extraction.DictVectorizer"><tt class="xref py py-class docutils literal"><span class="pre">feature_extraction.DictVectorizer</span></tt></a>) by <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a>.</li>
<li>Added Matthews correlation coefficient (<tt class="xref py py-func docutils literal"><span class="pre">metrics.matthews_corrcoef</span></tt>)
and added macro and micro average options to
<a class="reference internal" href="modules/generated/sklearn.metrics.precision_score.html#sklearn.metrics.precision_score" title="sklearn.metrics.precision_score"><tt class="xref py py-func docutils literal"><span class="pre">metrics.precision_score</span></tt></a>, <a class="reference internal" href="modules/generated/sklearn.metrics.recall_score.html#sklearn.metrics.recall_score" title="sklearn.metrics.recall_score"><tt class="xref py py-func docutils literal"><span class="pre">metrics.recall_score</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.metrics.f1_score.html#sklearn.metrics.f1_score" title="sklearn.metrics.f1_score"><tt class="xref py py-func docutils literal"><span class="pre">metrics.f1_score</span></tt></a> by <a class="reference external" href="http://www.mit.edu/~satra/">Satrajit Ghosh</a>.</li>
<li><a class="reference internal" href="modules/grid_search.html#out-of-bag"><em>Out of Bag Estimates</em></a> of generalization error for <a class="reference internal" href="modules/ensemble.html#ensemble"><em>Ensemble methods</em></a>
by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
<li><a class="reference internal" href="modules/feature_selection.html#randomized-l1"><em>Randomized sparse models</em></a>: Randomized sparse linear models for feature
selection, by <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a> and <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li><a class="reference internal" href="modules/label_propagation.html#label-propagation"><em>Label Propagation</em></a> for semi-supervised learning, by Clay
Woolam. <strong>Note</strong> the semi-supervised API is still work in progress,
and may change.</li>
<li>Added BIC/AIC model selection to classical <a class="reference internal" href="modules/mixture.html#gmm"><em>Gaussian mixture models</em></a> and unified
the API with the remainder of scikit-learn, by <a class="reference external" href="http://parietal.saclay.inria.fr/Members/bertrand-thirion">Bertrand Thirion</a></li>
<li>Added <a class="reference internal" href="modules/generated/sklearn.cross_validation.StratifiedShuffleSplit.html#sklearn.cross_validation.StratifiedShuffleSplit" title="sklearn.cross_validation.StratifiedShuffleSplit"><tt class="xref py py-class docutils literal"><span class="pre">sklearn.cross_validation.StratifiedShuffleSplit</span></tt></a>, which is
a <a class="reference internal" href="modules/generated/sklearn.cross_validation.ShuffleSplit.html#sklearn.cross_validation.ShuffleSplit" title="sklearn.cross_validation.ShuffleSplit"><tt class="xref py py-class docutils literal"><span class="pre">sklearn.cross_validation.ShuffleSplit</span></tt></a> with balanced splits,
by Yannick Schwartz.</li>
<li><a class="reference internal" href="modules/generated/sklearn.neighbors.NearestCentroid.html#sklearn.neighbors.NearestCentroid" title="sklearn.neighbors.NearestCentroid"><tt class="xref py py-class docutils literal"><span class="pre">sklearn.neighbors.NearestCentroid</span></tt></a> classifier added, along with a
<tt class="docutils literal"><span class="pre">shrink_threshold</span></tt> parameter, which implements <strong>shrunken centroid
classification</strong>, by <a class="reference external" href="http://www.twitter.com/robertlayton">Robert Layton</a>.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="other-changes">
<h3>Other changes<a class="headerlink" href="#other-changes" title="Permalink to this headline">¶</a></h3>
<blockquote>
<div><ul class="simple">
<li>Merged dense and sparse implementations of <a class="reference internal" href="modules/sgd.html#sgd"><em>Stochastic Gradient Descent</em></a> module and
exposed utility extension types for sequential
datasets <cite>seq_dataset</cite> and weight vectors <cite>weight_vector</cite>
by <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a>.</li>
<li>Added <cite>partial_fit</cite> (support for online/minibatch learning) and
warm_start to the <a class="reference internal" href="modules/sgd.html#sgd"><em>Stochastic Gradient Descent</em></a> module by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li>Dense and sparse implementations of <a class="reference internal" href="modules/svm.html#svm"><em>Support Vector Machines</em></a> classes and
<a class="reference internal" href="modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegression" title="sklearn.linear_model.LogisticRegression"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LogisticRegression</span></tt></a> merged by <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a>.</li>
<li>Regressors can now be used as base estimator in the <a class="reference internal" href="modules/multiclass.html#multiclass"><em>Multiclass and multilabel algorithms</em></a>
module by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li>Added n_jobs option to <a class="reference internal" href="modules/generated/sklearn.metrics.pairwise.pairwise_distances.html#sklearn.metrics.pairwise.pairwise_distances" title="sklearn.metrics.pairwise.pairwise_distances"><tt class="xref py py-func docutils literal"><span class="pre">metrics.pairwise.pairwise_distances</span></tt></a>
and <a class="reference internal" href="modules/generated/sklearn.metrics.pairwise.pairwise_kernels.html#sklearn.metrics.pairwise.pairwise_kernels" title="sklearn.metrics.pairwise.pairwise_kernels"><tt class="xref py py-func docutils literal"><span class="pre">metrics.pairwise.pairwise_kernels</span></tt></a> for parallel computation,
by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li><a class="reference internal" href="modules/clustering.html#k-means"><em>K-means</em></a> can now be run in parallel, using the <cite>n_jobs</cite> argument
to either <a class="reference internal" href="modules/clustering.html#k-means"><em>K-means</em></a> or <tt class="xref py py-class docutils literal"><span class="pre">KMeans</span></tt>, by <a class="reference external" href="http://www.twitter.com/robertlayton">Robert Layton</a>.</li>
<li>Improved <a class="reference internal" href="modules/cross_validation.html#cross-validation"><em>Cross-Validation: evaluating estimator performance</em></a> and <a class="reference internal" href="modules/grid_search.html#grid-search"><em>Grid Search: setting estimator parameters</em></a> documentation
and introduced the new <a class="reference internal" href="modules/generated/sklearn.cross_validation.train_test_split.html#sklearn.cross_validation.train_test_split" title="sklearn.cross_validation.train_test_split"><tt class="xref py py-func docutils literal"><span class="pre">cross_validation.train_test_split</span></tt></a>
helper function by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li><a class="reference internal" href="modules/generated/sklearn.svm.SVC.html#sklearn.svm.SVC" title="sklearn.svm.SVC"><tt class="xref py py-class docutils literal"><span class="pre">svm.SVC</span></tt></a> members <cite>coef_</cite> and <cite>intercept_</cite> changed sign for consistency
with <cite>decision_function</cite>; for <tt class="docutils literal"><span class="pre">kernel==linear</span></tt>, <cite>coef_</cite> was fixed
in the the one-vs-one case, by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
<li>Performance improvements to efficient leave-one-out cross-validated
Ridge regression, esp. for the <tt class="docutils literal"><span class="pre">n_samples</span> <span class="pre">></span> <span class="pre">n_features</span></tt> case, in
<a class="reference internal" href="modules/generated/sklearn.linear_model.RidgeCV.html#sklearn.linear_model.RidgeCV" title="sklearn.linear_model.RidgeCV"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.RidgeCV</span></tt></a>, by Reuben Fletcher-Costin.</li>
<li>Refactoring and simplication of the <a class="reference internal" href="modules/feature_extraction.html#text-feature-extraction"><em>Text feature extraction</em></a>
API and fixed a bug that caused possible negative IDF,
by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li>Beam pruning option in <tt class="xref py py-class docutils literal"><span class="pre">_BaseHMM</span></tt> 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.</li>
<li>Classes in <a class="reference internal" href="modules/neighbors.html#neighbors"><em>Nearest Neighbors</em></a> now support arbitrary Minkowski metric for
nearest neighbors searches. The metric can be specified by argument <tt class="docutils literal"><span class="pre">p</span></tt>.</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="id4">
<h2>API changes summary<a class="headerlink" href="#id4" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul>
<li><p class="first"><tt class="xref py py-class docutils literal"><span class="pre">covariance.EllipticEnvelop</span></tt> is now deprecated - Please use <a class="reference internal" href="modules/generated/sklearn.covariance.EllipticEnvelope.html#sklearn.covariance.EllipticEnvelope" title="sklearn.covariance.EllipticEnvelope"><tt class="xref py py-class docutils literal"><span class="pre">covariance.EllipticEnvelope</span></tt></a>
instead.</p>
</li>
<li><p class="first"><cite>NeighborsClassifier</cite> and <cite>NeighborsRegressor</cite> are gone in the module
<a class="reference internal" href="modules/neighbors.html#neighbors"><em>Nearest Neighbors</em></a>. Use the classes <tt class="xref py py-class docutils literal"><span class="pre">KNeighborsClassifier</span></tt>,
<tt class="xref py py-class docutils literal"><span class="pre">RadiusNeighborsClassifier</span></tt>, <tt class="xref py py-class docutils literal"><span class="pre">KNeighborsRegressor</span></tt>
and/or <tt class="xref py py-class docutils literal"><span class="pre">RadiusNeighborsRegressor</span></tt> instead.</p>
</li>
<li><p class="first">Sparse classes in the <a class="reference internal" href="modules/sgd.html#sgd"><em>Stochastic Gradient Descent</em></a> module are now deprecated.</p>
</li>
<li><p class="first">In <a class="reference internal" href="modules/generated/sklearn.mixture.GMM.html#sklearn.mixture.GMM" title="sklearn.mixture.GMM"><tt class="xref py py-class docutils literal"><span class="pre">mixture.GMM</span></tt></a>, <a class="reference internal" href="modules/generated/sklearn.mixture.DPGMM.html#sklearn.mixture.DPGMM" title="sklearn.mixture.DPGMM"><tt class="xref py py-class docutils literal"><span class="pre">mixture.DPGMM</span></tt></a> and <a class="reference internal" href="modules/generated/sklearn.mixture.VBGMM.html#sklearn.mixture.VBGMM" title="sklearn.mixture.VBGMM"><tt class="xref py py-class docutils literal"><span class="pre">mixture.VBGMM</span></tt></a>,
parameters must be passed to an object when initialising it and not through
<tt class="docutils literal"><span class="pre">fit</span></tt>. Now <tt class="docutils literal"><span class="pre">fit</span></tt> will only accept the data as an input parameter.</p>
</li>
<li><p class="first">methods <cite>rvs</cite> and <cite>decode</cite> in <tt class="xref py py-class docutils literal"><span class="pre">GMM</span></tt> module are now deprecated.
<cite>sample</cite> and <cite>score</cite> or <cite>predict</cite> should be used instead.</p>
</li>
<li><p class="first">attribute <cite>_scores</cite> and <cite>_pvalues</cite> in univariate feature selection
objects are now deprecated.
<cite>scores_</cite> or <cite>pvalues_</cite> should be used instead.</p>
</li>
<li><p class="first">In <tt class="xref py py-class docutils literal"><span class="pre">LogisticRegression</span></tt>, <tt class="xref py py-class docutils literal"><span class="pre">LinearSVC</span></tt>, <tt class="xref py py-class docutils literal"><span class="pre">SVC</span></tt> and
<tt class="xref py py-class docutils literal"><span class="pre">NuSVC</span></tt>, the <cite>class_weight</cite> parameter is now an initialization
parameter, not a parameter to fit. This makes grid searches
over this parameter possible.</p>
</li>
<li><p class="first">LFW <tt class="docutils literal"><span class="pre">data</span></tt> is now always shape <tt class="docutils literal"><span class="pre">(n_samples,</span> <span class="pre">n_features)</span></tt> to be
consistent with the Olivetti faces dataset. Use <tt class="docutils literal"><span class="pre">images</span></tt> and
<tt class="docutils literal"><span class="pre">pairs</span></tt> attribute to access the natural images shapes instead.</p>
</li>
<li><p class="first">In <a class="reference internal" href="modules/generated/sklearn.svm.LinearSVC.html#sklearn.svm.LinearSVC" title="sklearn.svm.LinearSVC"><tt class="xref py py-class docutils literal"><span class="pre">svm.LinearSVC</span></tt></a>, the meaning of the <cite>multi_class</cite> 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.</p>
</li>
<li><p class="first">Classs <tt class="xref py py-class docutils literal"><span class="pre">feature_selection.text.Vectorizer</span></tt> is deprecated and
replaced by <tt class="xref py py-class docutils literal"><span class="pre">feature_selection.text.TfidfVectorizer</span></tt>.</p>
</li>
<li><p class="first">The preprocessor / analyzer nested structure for text feature
extraction has been removed. All those features are
now directly passed as flat constructor arguments
to <tt class="xref py py-class docutils literal"><span class="pre">feature_selection.text.TfidfVectorizer</span></tt> and
<tt class="xref py py-class docutils literal"><span class="pre">feature_selection.text.CountVectorizer</span></tt>, in particular the
following parameters are now used:</p>
<blockquote>
<div><ul class="simple">
<li><tt class="docutils literal"><span class="pre">analyzer</span></tt> can be <cite>‘word’</cite> or <cite>‘char’</cite> to switch the default
analysis scheme, or use a specific python callable (as previously).</li>
<li><tt class="docutils literal"><span class="pre">tokenizer</span></tt> and <tt class="docutils literal"><span class="pre">preprocessor</span></tt> have been introduced to make it
still possible to customize those steps with the new API.</li>
<li><tt class="docutils literal"><span class="pre">input</span></tt> explicitly control how to interpret the sequence passed to
<tt class="docutils literal"><span class="pre">fit</span></tt> and <tt class="docutils literal"><span class="pre">predict</span></tt>: filenames, file objects or direct (byte or
unicode) strings.</li>
<li>charset decoding is explicit and strict by default.</li>
<li>the <tt class="docutils literal"><span class="pre">vocabulary</span></tt>, fitted or not is now stored in the
<tt class="docutils literal"><span class="pre">vocabulary_</span></tt> attribute to be consistent with the project
conventions.</li>
</ul>
</div></blockquote>
</li>
<li><p class="first">Class <tt class="xref py py-class docutils literal"><span class="pre">feature_selection.text.TfidfVectorizer</span></tt> now derives directly
from <tt class="xref py py-class docutils literal"><span class="pre">feature_selection.text.CountVectorizer</span></tt> to make grid
search trivial.</p>
</li>
<li><p class="first">methods <cite>rvs</cite> in <tt class="xref py py-class docutils literal"><span class="pre">_BaseHMM</span></tt> module are now deprecated.
<cite>sample</cite> should be used instead.</p>
</li>
<li><p class="first">Beam pruning option in <tt class="xref py py-class docutils literal"><span class="pre">_BaseHMM</span></tt> module is removed since it is
difficult to be Cythonized. If you are interested, you can look in the
history codes by git.</p>
</li>
<li><p class="first">The SVMlight format loader now supports files with both zero-based and
one-based column indices, since both occur “in the wild”.</p>
</li>
<li><p class="first">Arguments in class <tt class="xref py py-class docutils literal"><span class="pre">ShuffleSplit</span></tt> are now consistent with
<tt class="xref py py-class docutils literal"><span class="pre">StratifiedShuffleSplit</span></tt>. Arguments <tt class="docutils literal"><span class="pre">test_fraction</span></tt> and
<tt class="docutils literal"><span class="pre">train_fraction</span></tt> are deprecated and renamed to <tt class="docutils literal"><span class="pre">test_size</span></tt> and
<tt class="docutils literal"><span class="pre">train_size</span></tt> and can accept both <tt class="docutils literal"><span class="pre">float</span></tt> and <tt class="docutils literal"><span class="pre">int</span></tt>.</p>
</li>
<li><p class="first">Arguments in class <tt class="xref py py-class docutils literal"><span class="pre">Bootstrap</span></tt> are now consistent with
<tt class="xref py py-class docutils literal"><span class="pre">StratifiedShuffleSplit</span></tt>. Arguments <tt class="docutils literal"><span class="pre">n_test</span></tt> and
<tt class="docutils literal"><span class="pre">n_train</span></tt> are deprecated and renamed to <tt class="docutils literal"><span class="pre">test_size</span></tt> and
<tt class="docutils literal"><span class="pre">train_size</span></tt> and can accept both <tt class="docutils literal"><span class="pre">float</span></tt> and <tt class="docutils literal"><span class="pre">int</span></tt>.</p>
</li>
<li><p class="first">Argument <tt class="docutils literal"><span class="pre">p</span></tt> added to classes in <a class="reference internal" href="modules/neighbors.html#neighbors"><em>Nearest Neighbors</em></a> to specify an
arbitrary Minkowski metric for nearest neighbors searches.</p>
</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id5">
<h2>People<a class="headerlink" href="#id5" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>282 <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a></li>
<li>239 <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a></li>
<li>198 <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>129 <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>114 <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>103 Clay Woolam</li>
<li>96 <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a></li>
<li>88 <a class="reference external" href="https://github.com/jaquesgrobler/scikit-learn/wiki/Jaques-Grobler">Jaques Grobler</a></li>
<li>82 <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>50 <a class="reference external" href="http://parietal.saclay.inria.fr/Members/bertrand-thirion">Bertrand Thirion</a></li>
<li>42 <a class="reference external" href="http://www.twitter.com/robertlayton">Robert Layton</a></li>
<li>28 flyingimmidev</li>
<li>26 <a class="reference external" href="http://www.astro.washington.edu/users/vanderplas/">Jake Vanderplas</a></li>
<li>26 Shiqiao Du</li>
<li>21 <a class="reference external" href="http://www.mit.edu/~satra/">Satrajit Ghosh</a></li>
<li>17 <a class="reference external" href="http://www.davidmarek.cz/">David Marek</a></li>
<li>17 <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a></li>
<li>14 <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>11 Yannick Schwartz</li>
<li>10 <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a></li>
<li>9 fcostin</li>
<li>7 Nick Wilson</li>
<li>5 Adrien Gaidon</li>
<li>5 <a class="reference external" href="http://pinto.scripts.mit.edu/">Nicolas Pinto</a></li>
<li>4 <a class="reference external" href="http://www-etud.iro.umontreal.ca/~wardefar/">David Warde-Farley</a></li>
<li>5 Nelle Varoquaux</li>
<li>5 Emmanuelle Gouillart</li>
<li>3 Joonas Sillanpää</li>
<li>3 Paolo Losi</li>
<li>2 Charles McCarthy</li>
<li>2 Roy Hyunjin Han</li>
<li>2 Scott White</li>
<li>2 ibayer</li>
<li>1 Brandyn White</li>
<li>1 Carlos Scheidegger</li>
<li>1 Claire Revillet</li>
<li>1 Conrad Lee</li>
<li>1 <a class="reference external" href="http://www.lnao.fr/spip.php?rubrique30">Edouard Duchesnay</a></li>
<li>1 Jan Hendrik Metzen</li>
<li>1 Meng Xinfan</li>
<li>1 Rob Zinkov</li>
<li>1 Shiqiao</li>
<li>1 Udi Weinsberg</li>
<li>1 Virgile Fritsch</li>
<li>1 Xinfan Meng</li>
<li>1 Yaroslav Halchenko</li>
<li>1 jansoe</li>
<li>1 Leon Palafox</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="changes-0-10">
<span id="id6"></span><h1>0.10<a class="headerlink" href="#changes-0-10" title="Permalink to this headline">¶</a></h1>
<div class="section" id="id7">
<h2>Changelog<a class="headerlink" href="#id7" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>Python 2.5 compatibility was dropped; the minimum Python version needed
to use scikit-learn is now 2.6.</li>
<li><a class="reference internal" href="modules/covariance.html#sparse-inverse-covariance"><em>Sparse inverse covariance</em></a> estimation using the graph Lasso, with
associated cross-validated estimator, by <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>New <a class="reference internal" href="modules/tree.html#tree"><em>Tree</em></a> module by <a class="reference external" href="http://info.ee.surrey.ac.uk/Personal/B.Holt/">Brian Holt</a>, <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a>,
<a class="reference external" href="http://www.mit.edu/~satra/">Satrajit Ghosh</a> and <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a>. The module comes with complete
documentation and examples.</li>
<li>Fixed a bug in the RFE module by <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a> (issue #378).</li>
<li>Fixed a memory leak in in <a class="reference internal" href="modules/svm.html#svm"><em>Support Vector Machines</em></a> module by <a class="reference external" href="http://info.ee.surrey.ac.uk/Personal/B.Holt/">Brian Holt</a> (issue #367).</li>
<li>Faster tests by <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a> and others.</li>
<li>Silhouette Coefficient cluster analysis evaluation metric added as
<a class="reference internal" href="modules/generated/sklearn.metrics.silhouette_score.html#sklearn.metrics.silhouette_score" title="sklearn.metrics.silhouette_score"><tt class="xref py py-func docutils literal"><span class="pre">sklearn.metrics.silhouette_score</span></tt></a> by Robert Layton.</li>
<li>Fixed a bug in <a class="reference internal" href="modules/clustering.html#k-means"><em>K-means</em></a> in the handling of the <tt class="docutils literal"><span class="pre">n_init</span></tt> parameter:
the clustering algorithm used to be run <tt class="docutils literal"><span class="pre">n_init</span></tt> times but the last
solution was retained instead of the best solution by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li>Minor refactoring in <a class="reference internal" href="modules/sgd.html#sgd"><em>Stochastic Gradient Descent</em></a> module; consolidated dense and sparse
predict methods; Enhanced test time performance by converting model
parameters to fortran-style arrays after fitting (only multi-class).</li>
<li>Adjusted Mutual Information metric added as
<a class="reference internal" href="modules/generated/sklearn.metrics.adjusted_mutual_info_score.html#sklearn.metrics.adjusted_mutual_info_score" title="sklearn.metrics.adjusted_mutual_info_score"><tt class="xref py py-func docutils literal"><span class="pre">sklearn.metrics.adjusted_mutual_info_score</span></tt></a> by Robert Layton.</li>
<li>Models like SVC/SVR/LinearSVC/LogisticRegression from libsvm/liblinear
now support scaling of C regularization parameter by the number of
samples by <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a>.</li>
<li>New <a class="reference internal" href="modules/ensemble.html#ensemble"><em>Ensemble Methods</em></a> module by <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a> and
<a class="reference external" href="http://info.ee.surrey.ac.uk/Personal/B.Holt/">Brian Holt</a>. The module comes with the random forest algorithm and the
extra-trees method, along with documentation and examples.</li>
<li><a class="reference internal" href="modules/outlier_detection.html#outlier-detection"><em>Novelty and Outlier Detection</em></a>: outlier and novelty detection, by
<a class="reference external" href="http://parietal.saclay.inria.fr/Members/virgile-fritsch">Virgile Fritsch</a>.</li>
<li><a class="reference internal" href="modules/kernel_approximation.html#kernel-approximation"><em>Kernel Approximation</em></a>: a transform implementing kernel
approximation for fast SGD on non-linear kernels by
<a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
<li>Fixed a bug due to atom swapping in <a class="reference internal" href="modules/linear_model.html#omp"><em>Orthogonal Matching Pursuit (OMP)</em></a> by <a class="reference external" href="http://vene.ro">Vlad Niculae</a>.</li>
<li><a class="reference internal" href="modules/decomposition.html#sparsecoder"><em>Sparse coding with a precomputed dictionary</em></a> by <a class="reference external" href="http://vene.ro">Vlad Niculae</a>.</li>
<li><a class="reference internal" href="modules/clustering.html#mini-batch-kmeans"><em>Mini Batch K-Means</em></a> performance improvements by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li><a class="reference internal" href="modules/clustering.html#k-means"><em>K-means</em></a> support for sparse matrices by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li>Improved documentation for developers and for the <a class="reference internal" href="modules/classes.html#module-sklearn.utils" title="sklearn.utils"><tt class="xref py py-mod docutils literal"><span class="pre">sklearn.utils</span></tt></a>
module, by <a class="reference external" href="http://www.astro.washington.edu/users/vanderplas/">Jake VanderPlas</a>.</li>
<li>Vectorized 20newsgroups dataset loader
(<a class="reference internal" href="modules/generated/sklearn.datasets.fetch_20newsgroups_vectorized.html#sklearn.datasets.fetch_20newsgroups_vectorized" title="sklearn.datasets.fetch_20newsgroups_vectorized"><tt class="xref py py-func docutils literal"><span class="pre">sklearn.datasets.fetch_20newsgroups_vectorized</span></tt></a>) by
<a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li><a class="reference internal" href="modules/multiclass.html#multiclass"><em>Multiclass and multilabel algorithms</em></a> by <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a>.</li>
<li>Utilities for fast computation of mean and variance for sparse matrices
by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li>Make <a class="reference internal" href="modules/generated/sklearn.preprocessing.scale.html#sklearn.preprocessing.scale" title="sklearn.preprocessing.scale"><tt class="xref py py-func docutils literal"><span class="pre">sklearn.preprocessing.scale</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.preprocessing.Scaler.html#sklearn.preprocessing.Scaler" title="sklearn.preprocessing.Scaler"><tt class="xref py py-class docutils literal"><span class="pre">sklearn.preprocessing.Scaler</span></tt></a> work on sparse matrices by
<a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>Feature importances using decision trees and/or forest of trees,
by <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a>.</li>
<li>Parallel implementation of forests of randomized trees by
<a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a>.</li>
<li><a class="reference internal" href="modules/generated/sklearn.cross_validation.ShuffleSplit.html#sklearn.cross_validation.ShuffleSplit" title="sklearn.cross_validation.ShuffleSplit"><tt class="xref py py-class docutils literal"><span class="pre">sklearn.cross_validation.ShuffleSplit</span></tt></a> can subsample the train
sets as well as the test sets by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li>Errors in the build of the documentation fixed by <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a>.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id8">
<h2>API changes summary<a class="headerlink" href="#id8" title="Permalink to this headline">¶</a></h2>
<p>Here are the code migration instructions when updgrading from scikit-learn
version 0.9:</p>
<blockquote>
<div><ul>
<li><p class="first">Some estimators that may overwrite their inputs to save memory previously
had <tt class="docutils literal"><span class="pre">overwrite_</span></tt> parameters; these have been replaced with <tt class="docutils literal"><span class="pre">copy_</span></tt>
parameters with exactly the opposite meaning.</p>
<p>This particularly affects some of the estimators in <tt class="xref py py-mod docutils literal"><span class="pre">linear_model</span></tt>.
The default behavior is still to copy everything passed in.</p>
</li>
<li><p class="first">The SVMlight dataset loader <a class="reference internal" href="modules/generated/sklearn.datasets.load_svmlight_file.html#sklearn.datasets.load_svmlight_file" title="sklearn.datasets.load_svmlight_file"><tt class="xref py py-func docutils literal"><span class="pre">sklearn.datasets.load_svmlight_file</span></tt></a> no
longer supports loading two files at once; use <tt class="docutils literal"><span class="pre">load_svmlight_files</span></tt>
instead. Also, the (unused) <tt class="docutils literal"><span class="pre">buffer_mb</span></tt> parameter is gone.</p>
</li>
<li><p class="first">Sparse estimators in the <a class="reference internal" href="modules/sgd.html#sgd"><em>Stochastic Gradient Descent</em></a> module use dense parameter vector
<tt class="docutils literal"><span class="pre">coef_</span></tt> instead of <tt class="docutils literal"><span class="pre">sparse_coef_</span></tt>. This significantly improves
test time performance.</p>
</li>
<li><p class="first">The <a class="reference internal" href="modules/covariance.html#covariance"><em>Covariance estimation</em></a> module now has a robust estimator of
covariance, the Minimum Covariance Determinant estimator.</p>
</li>
<li><p class="first">Cluster evaluation metrics in <tt class="xref py py-mod docutils literal"><span class="pre">metrics.cluster</span></tt> have been refactored
but the changes are backwards compatible. They have been moved to the
<tt class="xref py py-mod docutils literal"><span class="pre">metrics.cluster.supervised</span></tt>, along with
<tt class="xref py py-mod docutils literal"><span class="pre">metrics.cluster.unsupervised</span></tt> which contains the Silhouette
Coefficient.</p>
</li>
<li><p class="first">The <tt class="docutils literal"><span class="pre">permutation_test_score</span></tt> function now behaves the same way as
<tt class="docutils literal"><span class="pre">cross_val_score</span></tt> (i.e. uses the mean score across the folds.)</p>
</li>
<li><p class="first">Cross Validation generators now use integer indices (<tt class="docutils literal"><span class="pre">indices=True</span></tt>)
by default instead of boolean masks. This make it more intuitive to
use with sparse matrix data.</p>
</li>
<li><p class="first">The functions used for sparse coding, <tt class="docutils literal"><span class="pre">sparse_encode</span></tt> and
<tt class="docutils literal"><span class="pre">sparse_encode_parallel</span></tt> have been combined into
<a class="reference internal" href="modules/generated/sklearn.decomposition.sparse_encode.html#sklearn.decomposition.sparse_encode" title="sklearn.decomposition.sparse_encode"><tt class="xref py py-func docutils literal"><span class="pre">sklearn.decomposition.sparse_encode</span></tt></a>, and the shapes of the arrays
have been transposed for consistency with the matrix factorization setting,
as opposed to the regression setting.</p>
</li>
<li><p class="first">Fixed an off-by-one error in the SVMlight/LibSVM file format handling;
files generated using <tt class="xref py py-func docutils literal"><span class="pre">sklearn.datasets.dump_svmlight_file</span></tt> should be
re-generated. (They should continue to work, but accidentally had one
extra column of zeros prepended.)</p>
</li>
<li><p class="first"><tt class="docutils literal"><span class="pre">BaseDictionaryLearning</span></tt> class replaced by <tt class="docutils literal"><span class="pre">SparseCodingMixin</span></tt>.</p>
</li>
<li><p class="first"><tt class="xref py py-func docutils literal"><span class="pre">sklearn.utils.extmath.fast_svd</span></tt> has been renamed
<tt class="xref py py-func docutils literal"><span class="pre">sklearn.utils.extmath.randomized_svd</span></tt> 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.</p>
</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id9">
<h2>People<a class="headerlink" href="#id9" title="Permalink to this headline">¶</a></h2>
<p>The following people contributed to scikit-learn since last release:</p>
<blockquote>
<div><ul class="simple">
<li>246 <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a></li>
<li>242 <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>220 <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a></li>
<li>183 <a class="reference external" href="http://info.ee.surrey.ac.uk/Personal/B.Holt/">Brian Holt</a></li>
<li>166 <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>144 <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a></li>
<li>73 <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>65 <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a></li>
<li>64 <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a></li>
<li>60 Robert Layton</li>
<li>55 <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>52 <a class="reference external" href="http://www.astro.washington.edu/users/vanderplas/">Jake Vanderplas</a></li>
<li>44 Noel Dawe</li>
<li>38 <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>24 <a class="reference external" href="http://parietal.saclay.inria.fr/Members/virgile-fritsch">Virgile Fritsch</a></li>
<li>23 <a class="reference external" href="http://www.mit.edu/~satra/">Satrajit Ghosh</a></li>
<li>3 Jan Hendrik Metzen</li>
<li>3 Kenneth C. Arnold</li>
<li>3 Shiqiao Du</li>
<li>3 Tim Sheerman-Chase</li>
<li>3 <a class="reference external" href="http://www.onerussian.com/">Yaroslav Halchenko</a></li>
<li>2 Bala Subrahmanyam Varanasi</li>
<li>2 DraXus</li>
<li>2 Michael Eickenberg</li>
<li>1 Bogdan Trach</li>
<li>1 Félix-Antoine Fortin</li>
<li>1 Juan Manuel Caicedo Carvajal</li>
<li>1 Nelle Varoquaux</li>
<li>1 <a class="reference external" href="http://pinto.scripts.mit.edu/">Nicolas Pinto</a></li>
<li>1 Tiziano Zito</li>
<li>1 Xinfan Meng</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="changes-0-9">
<span id="id10"></span><h1>0.9<a class="headerlink" href="#changes-0-9" title="Permalink to this headline">¶</a></h1>
<p>scikit-learn 0.9 was released on September 2011, three months after the 0.8
release and includes the new modules <a class="reference internal" href="modules/manifold.html#manifold"><em>Manifold learning</em></a>, <a class="reference internal" href="modules/mixture.html#dirichlet-process"><em>The Dirichlet Process</em></a>
as well as several new algorithms and documentation improvements.</p>
<p>This release also includes the dictionary-learning work developed by
<a class="reference external" href="http://vene.ro">Vlad Niculae</a> as part of the <a class="reference external" href="http://code.google.com/soc/">Google Summer of Code</a> program.</p>
<p><div style="text-align: center; margin: 0px 0 -5px 0;"> <a class="reference external" href="auto_examples/linear_model/plot_omp.html"><img alt="banner2" src="_images/plot_omp1.png" /></a> <a class="reference external" href="auto_examples/manifold/plot_compare_methods.html"><img alt="banner1" src="_images/plot_compare_methods1.png" /></a> <a class="reference external" href="auto_examples/decomposition/plot_kernel_pca.html"><img alt="banner3" src="_images/plot_kernel_pca1.png" /></a> </div></p>
<div class="section" id="id11">
<h2>Changelog<a class="headerlink" href="#id11" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>New <a class="reference internal" href="modules/manifold.html#manifold"><em>Manifold learning</em></a> module by <a class="reference external" href="http://www.astro.washington.edu/users/vanderplas/">Jake Vanderplas</a> and
<a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>.</li>
<li>New <a class="reference internal" href="modules/mixture.html#dirichlet-process"><em>Dirichlet Process</em></a> Gaussian Mixture
Model by <a class="reference external" href="<http://atpassos.posterous.com>">Alexandre Passos</a></li>
<li><a class="reference internal" href="modules/neighbors.html#neighbors"><em>Nearest Neighbors</em></a> module refactoring by <a class="reference external" href="http://www.astro.washington.edu/users/vanderplas/">Jake Vanderplas</a> :
general refactoring, support for sparse matrices in input, speed and
documentation improvements. See the next section for a full list of API
changes.</li>
<li>Improvements on the <a class="reference internal" href="modules/feature_selection.html#feature-selection"><em>Feature selection</em></a> module by
<a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a> : refactoring of the RFE classes, documentation
rewrite, increased efficiency and minor API changes.</li>
<li><a class="reference internal" href="modules/decomposition.html#sparsepca"><em>Sparse Principal Components Analysis (SparsePCA and MiniBatchSparsePCA)</em></a> by <a class="reference external" href="http://vene.ro">Vlad Niculae</a>, <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a> and
<a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>Printing an estimator now behaves independently of architectures
and Python version thanks to Jean Kossaifi.</li>
<li><a class="reference internal" href="datasets/index.html#libsvm-loader"><em>Loader for libsvm/svmlight format</em></a> by
<a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a> and <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a></li>
<li>Documentation improvements: thumbnails in
<a class="reference internal" href="auto_examples/index.html#examples-index"><em>example gallery</em></a> by <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>.</li>
<li>Important bugfixes in <a class="reference internal" href="modules/svm.html#svm"><em>Support Vector Machines</em></a> module (segfaults, bad
performance) by <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>.</li>
<li>Added <a class="reference internal" href="modules/naive_bayes.html#multinomial-naive-bayes"><em>Multinomial Naive Bayes</em></a> and <a class="reference internal" href="modules/naive_bayes.html#bernoulli-naive-bayes"><em>Bernoulli Naive Bayes</em></a>
by <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a></li>
<li>Text feature extraction optimizations by Lars Buitinck</li>
<li>Chi-Square feature selection
(<tt class="xref py py-func docutils literal"><span class="pre">feature_selection.univariate_selection.chi2</span></tt>) by <cite>Lars Buitinck</cite>.</li>
<li><a class="reference internal" href="datasets/index.html#sample-generators"><em>Sample generators</em></a> module refactoring by <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a></li>
<li><a class="reference internal" href="modules/multiclass.html#multiclass"><em>Multiclass and multilabel algorithms</em></a> by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>Ball tree rewrite by <a class="reference external" href="http://www.astro.washington.edu/users/vanderplas/">Jake Vanderplas</a></li>
<li>Implementation of <a class="reference internal" href="modules/clustering.html#dbscan"><em>DBSCAN</em></a> algorithm by Robert Layton</li>
<li>Kmeans predict and transform by Robert Layton</li>
<li>Preprocessing module refactoring by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>Faster mean shift by Conrad Lee</li>
<li>New <a class="reference internal" href="modules/cross_validation.html#bootstrap"><em>Bootstrapping cross-validation</em></a>, <a class="reference internal" href="modules/cross_validation.html#shufflesplit"><em>Random permutations cross-validation a.k.a. Shuffle & Split</em></a> and various other
improvements in cross validation schemes by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a> and
<a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>Adjusted Rand index and V-Measure clustering evaluation metrics by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>Added <a class="reference internal" href="modules/generated/sklearn.linear_model.OrthogonalMatchingPursuit.html#sklearn.linear_model.OrthogonalMatchingPursuit" title="sklearn.linear_model.OrthogonalMatchingPursuit"><tt class="xref py py-class docutils literal"><span class="pre">Orthogonal</span> <span class="pre">Matching</span> <span class="pre">Pursuit</span></tt></a> by <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>Added 2D-patch extractor utilites in the <a class="reference internal" href="modules/feature_extraction.html#feature-extraction"><em>Feature extraction</em></a> module by <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>Implementation of <a class="reference internal" href="modules/generated/sklearn.linear_model.LassoLarsCV.html#sklearn.linear_model.LassoLarsCV" title="sklearn.linear_model.LassoLarsCV"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LassoLarsCV</span></tt></a>
(cross-validated Lasso solver using the Lars algorithm) and
<a class="reference internal" href="modules/generated/sklearn.linear_model.LassoLarsIC.html#sklearn.linear_model.LassoLarsIC" title="sklearn.linear_model.LassoLarsIC"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LassoLarsIC</span></tt></a> (BIC/AIC model
selection in Lars) by <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a>
and <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>Scalability improvements to <a class="reference internal" href="modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve" title="sklearn.metrics.roc_curve"><tt class="xref py py-func docutils literal"><span class="pre">metrics.roc_curve</span></tt></a> by Olivier Hervieu</li>
<li>Distance helper functions <a class="reference internal" href="modules/generated/sklearn.metrics.pairwise.pairwise_distances.html#sklearn.metrics.pairwise.pairwise_distances" title="sklearn.metrics.pairwise.pairwise_distances"><tt class="xref py py-func docutils literal"><span class="pre">metrics.pairwise.pairwise_distances</span></tt></a>
and <a class="reference internal" href="modules/generated/sklearn.metrics.pairwise.pairwise_kernels.html#sklearn.metrics.pairwise.pairwise_kernels" title="sklearn.metrics.pairwise.pairwise_kernels"><tt class="xref py py-func docutils literal"><span class="pre">metrics.pairwise.pairwise_kernels</span></tt></a> by Robert Layton</li>
<li><a class="reference internal" href="modules/generated/sklearn.cluster.MiniBatchKMeans.html#sklearn.cluster.MiniBatchKMeans" title="sklearn.cluster.MiniBatchKMeans"><tt class="xref py py-class docutils literal"><span class="pre">Mini-Batch</span> <span class="pre">K-Means</span></tt></a> by Nelle Varoquaux and Peter Prettenhofer.</li>
<li><a class="reference internal" href="datasets/index.html#mldata"><em>Downloading datasets from the mldata.org repository</em></a> utilities by Pietro Berkes.</li>
<li><a class="reference internal" href="datasets/index.html#olivetti-faces"><em>The Olivetti faces dataset</em></a> by <a class="reference external" href="http://www-etud.iro.umontreal.ca/~wardefar/">David Warde-Farley</a>.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id12">
<h2>API changes summary<a class="headerlink" href="#id12" title="Permalink to this headline">¶</a></h2>
<p>Here are the code migration instructions when updgrading from scikit-learn
version 0.8:</p>
<blockquote>
<div><ul>
<li><p class="first">The <tt class="docutils literal"><span class="pre">scikits.learn</span></tt> package was renamed <tt class="docutils literal"><span class="pre">sklearn</span></tt>. There is
still a <tt class="docutils literal"><span class="pre">scikits.learn</span></tt> package alias for backward compatibility.</p>
<p>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!):</p>
<div class="highlight-python"><pre>find -name "*.py" | xargs sed -i 's/\bscikits.learn\b/sklearn/g'</pre>
</div>
</li>
<li><p class="first">Estimators no longer accept model parameters as <tt class="docutils literal"><span class="pre">fit</span></tt> arguments:
instead all parameters must be only be passed as constructor
arguments or using the now public <tt class="docutils literal"><span class="pre">set_params</span></tt> method inhereted
from <tt class="xref py py-class docutils literal"><span class="pre">base.BaseEstimator</span></tt>.</p>
<p>Some estimators can still accept keyword arguments on the <tt class="docutils literal"><span class="pre">fit</span></tt>
but this is restricted to data-dependent values (e.g. a Gram matrix
or an affinity matrix that are precomputed from the <tt class="docutils literal"><span class="pre">X</span></tt> data matrix.</p>
</li>
<li><p class="first">The <tt class="docutils literal"><span class="pre">cross_val</span></tt> package has been renamed to <tt class="docutils literal"><span class="pre">cross_validation</span></tt>
although there is also a <tt class="docutils literal"><span class="pre">cross_val</span></tt> package alias in place for
backward compatibility.</p>
<p>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!):</p>
<div class="highlight-python"><pre>find -name "*.py" | xargs sed -i 's/\bcross_val\b/cross_validation/g'</pre>
</div>
</li>
<li><p class="first">The <tt class="docutils literal"><span class="pre">score_func</span></tt> argument of the
<tt class="docutils literal"><span class="pre">sklearn.cross_validation.cross_val_score</span></tt> function is now expected
to accept <tt class="docutils literal"><span class="pre">y_test</span></tt> and <tt class="docutils literal"><span class="pre">y_predicted</span></tt> as only arguments for
classification and regression tasks or <tt class="docutils literal"><span class="pre">X_test</span></tt> for unsupervised
estimators.</p>
</li>
<li><p class="first"><tt class="docutils literal"><span class="pre">gamma</span></tt> parameter for support vector machine algorithms is set
to <tt class="docutils literal"><span class="pre">1</span> <span class="pre">/</span> <span class="pre">n_features</span></tt> by default, instead of <tt class="docutils literal"><span class="pre">1</span> <span class="pre">/</span> <span class="pre">n_samples</span></tt>.</p>
</li>
<li><p class="first">The <tt class="docutils literal"><span class="pre">sklearn.hmm</span></tt> 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.</p>
</li>
<li><p class="first"><tt class="docutils literal"><span class="pre">sklearn.neighbors</span></tt> has been made into a submodule. The two previously
available estimators, <tt class="docutils literal"><span class="pre">NeighborsClassifier</span></tt> and <tt class="docutils literal"><span class="pre">NeighborsRegressor</span></tt>
have been marked as deprecated. Their functionality has been divided
among five new classes: <tt class="docutils literal"><span class="pre">NearestNeighbors</span></tt> for unsupervised neighbors
searches, <tt class="docutils literal"><span class="pre">KNeighborsClassifier</span></tt> & <tt class="docutils literal"><span class="pre">RadiusNeighborsClassifier</span></tt>
for supervised classification problems, and <tt class="docutils literal"><span class="pre">KNeighborsRegressor</span></tt>
& <tt class="docutils literal"><span class="pre">RadiusNeighborsRegressor</span></tt> for supervised regression problems.</p>
</li>
<li><p class="first"><tt class="docutils literal"><span class="pre">sklearn.ball_tree.BallTree</span></tt> has been moved to
<tt class="docutils literal"><span class="pre">sklearn.neighbors.BallTree</span></tt>. Using the former will generate a warning.</p>
</li>
<li><p class="first"><tt class="docutils literal"><span class="pre">sklearn.linear_model.LARS()</span></tt> and related classes (LassoLARS,
LassoLARSCV, etc.) have been renamed to
<tt class="docutils literal"><span class="pre">sklearn.linear_model.Lars()</span></tt>.</p>
</li>
<li><p class="first">All distance metrics and kernels in <tt class="docutils literal"><span class="pre">sklearn.metrics.pairwise</span></tt> 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.</p>
</li>
<li><p class="first"><tt class="docutils literal"><span class="pre">sklearn.metrics.pairwise.l1_distance</span></tt> is now called <tt class="docutils literal"><span class="pre">manhattan_distance</span></tt>,
and by default returns the pairwise distance. For the component wise distance,
set the parameter <tt class="docutils literal"><span class="pre">sum_over_features</span></tt> to <tt class="docutils literal"><span class="pre">False</span></tt>.</p>
</li>
</ul>
</div></blockquote>
<p>Backward compatibilty package aliases and other deprecated classes and
functions will be removed in version 0.11.</p>
</div>
<div class="section" id="id13">
<h2>People<a class="headerlink" href="#id13" title="Permalink to this headline">¶</a></h2>
<p>38 people contributed to this release.</p>
<blockquote>
<div><ul class="simple">
<li>387 <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>320 <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>192 <a class="reference external" href="https://github.com/larsmans">Lars Buitinck</a></li>
<li>179 <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>168 <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a> (<a class="reference external" href="http://inria.fr">INRIA</a>, <a class="reference external" href="http://parietal.saclay.inria.fr/">Parietal Team</a>)</li>
<li>127 <a class="reference external" href="http://www.astro.washington.edu/users/vanderplas/">Jake Vanderplas</a></li>
<li>120 <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>85 <a class="reference external" href="<http://atpassos.posterous.com>">Alexandre Passos</a></li>
<li>67 <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>57 <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a></li>
<li>56 <a class="reference external" href="http://www.montefiore.ulg.ac.be/~glouppe/">Gilles Louppe</a></li>
<li>42 Robert Layton</li>
<li>38 Nelle Varoquaux</li>
<li>32 Jean Kossaifi</li>
<li>30 Conrad Lee</li>
<li>22 Pietro Berkes</li>
<li>18 andy</li>
<li>17 David Warde-Farley</li>
<li>12 Brian Holt</li>
<li>11 Robert</li>
<li>8 Amit Aides</li>
<li>8 <a class="reference external" href="http://parietal.saclay.inria.fr/Members/virgile-fritsch">Virgile Fritsch</a></li>
<li>7 <a class="reference external" href="http://www.onerussian.com/">Yaroslav Halchenko</a></li>
<li>6 Salvatore Masecchia</li>
<li>5 Paolo Losi</li>
<li>4 Vincent Schut</li>
<li>3 Alexis Metaireau</li>
<li>3 Bryan Silverthorn</li>
<li>3 <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a></li>
<li>2 Minwoo Jake Lee</li>
<li>1 Emmanuelle Gouillart</li>
<li>1 Keith Goodman</li>
<li>1 Lucas Wiman</li>
<li>1 <a class="reference external" href="http://pinto.scripts.mit.edu/">Nicolas Pinto</a></li>
<li>1 Thouis (Ray) Jones</li>
<li>1 Tim Sheerman-Chase</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="changes-0-8">
<span id="id14"></span><h1>0.8<a class="headerlink" href="#changes-0-8" title="Permalink to this headline">¶</a></h1>
<p>scikit-learn 0.8 was released on May 2011, one month after the first
“international” <a class="reference external" href="https://github.com/scikit-learn/scikit-learn/wiki/Upcoming-events">scikit-learn coding sprint</a> and is
marked by the inclusion of important modules: <a class="reference internal" href="modules/clustering.html#hierarchical-clustering"><em>Hierarchical clustering</em></a>,
<a class="reference internal" href="modules/pls.html#pls"><em>Partial Least Squares</em></a>, <a class="reference internal" href="modules/decomposition.html#nmf"><em>Non-negative matrix factorization (NMF or NNMF)</em></a>, initial support for Python 3 and by important
enhacements and bug fixes.</p>
<div class="section" id="id15">
<h2>Changelog<a class="headerlink" href="#id15" title="Permalink to this headline">¶</a></h2>
<p>Several new modules where introduced during this release:</p>
<blockquote>
<div><ul class="simple">
<li>New <a class="reference internal" href="modules/clustering.html#hierarchical-clustering"><em>Hierarchical clustering</em></a> module by Vincent Michel,
<a class="reference external" href="http://parietal.saclay.inria.fr/Members/bertrand-thirion">Bertrand Thirion</a>, <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a> and <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a>.</li>
<li><a class="reference internal" href="modules/decomposition.html#kernel-pca"><em>Kernel PCA</em></a> implementation by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li><a class="reference internal" href="datasets/index.html#labeled-faces-in-the-wild"><em>The Labeled Faces in the Wild face recognition dataset</em></a> by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li>New <a class="reference internal" href="modules/pls.html#pls"><em>Partial Least Squares</em></a> module by <a class="reference external" href="http://www.lnao.fr/spip.php?rubrique30">Edouard Duchesnay</a>.</li>
<li><a class="reference internal" href="modules/decomposition.html#nmf"><em>Non-negative matrix factorization (NMF or NNMF)</em></a> module <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>Implementation of the <a class="reference internal" href="modules/covariance.html#oracle-approximating-shrinkage"><em>Oracle Approximating Shrinkage</em></a> algorithm by
<a class="reference external" href="http://parietal.saclay.inria.fr/Members/virgile-fritsch">Virgile Fritsch</a> in the <a class="reference internal" href="modules/covariance.html#covariance"><em>Covariance estimation</em></a> module.</li>
</ul>
</div></blockquote>
<p>Some other modules benefited from significant improvements or cleanups.</p>
<blockquote>
<div><ul class="simple">
<li>Initial support for Python 3: builds and imports cleanly,
some modules are usable while others have failing tests by <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>.</li>
<li><a class="reference internal" href="modules/generated/sklearn.decomposition.PCA.html#sklearn.decomposition.PCA" title="sklearn.decomposition.PCA"><tt class="xref py py-class docutils literal"><span class="pre">decomposition.PCA</span></tt></a> is now usable from the Pipeline object by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li>Guide <a class="reference internal" href="developers/performance.html#performance-howto"><em>How to optimize for speed</em></a> by <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>.</li>
<li>Fixes for memory leaks in libsvm bindings, 64-bit safer BallTree by Lars Buitinck.</li>
<li>bug and style fixing in <a class="reference internal" href="modules/clustering.html#k-means"><em>K-means</em></a> algorithm by Jan Schlüter.</li>
<li>Add attribute coverged to Gaussian Mixture Models by Vincent Schut.</li>
<li>Implement <cite>transform</cite>, <cite>predict_log_proba</cite> in <a class="reference internal" href="modules/generated/sklearn.lda.LDA.html#sklearn.lda.LDA" title="sklearn.lda.LDA"><tt class="xref py py-class docutils literal"><span class="pre">lda.LDA</span></tt></a> by <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>.</li>
<li>Refactoring in the <a class="reference internal" href="modules/svm.html#svm"><em>Support Vector Machines</em></a> module and bug fixes by <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>,
<a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a> and Amit Aides.</li>
<li>Refactored SGD module (removed code duplication, better variable naming),
added interface for sample weight by <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a>.</li>
<li>Wrapped BallTree with Cython by Thouis (Ray) Jones.</li>
<li>Added function <a class="reference internal" href="modules/generated/sklearn.svm.l1_min_c.html#sklearn.svm.l1_min_c" title="sklearn.svm.l1_min_c"><tt class="xref py py-func docutils literal"><span class="pre">svm.l1_min_c</span></tt></a> by Paolo Losi.</li>
<li>Typos, doc style, etc. by <a class="reference external" href="http://www.onerussian.com/">Yaroslav Halchenko</a>, <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a>,
<a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>, Yann Malet, <a class="reference external" href="http://pinto.scripts.mit.edu/">Nicolas Pinto</a>, Lars Buitinck and
<a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id16">
<h2>People<a class="headerlink" href="#id16" title="Permalink to this headline">¶</a></h2>
<p>People that made this release possible preceeded by number of commits:</p>
<blockquote>
<div><ul class="simple">
<li>159 <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>96 <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>96 <a class="reference external" href="http://vene.ro">Vlad Niculae</a></li>
<li>94 <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a></li>
<li>36 <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>32 Paolo Losi</li>
<li>31 <a class="reference external" href="http://www.lnao.fr/spip.php?rubrique30">Edouard Duchesnay</a></li>
<li>30 <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>25 <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a></li>
<li>22 <a class="reference external" href="http://pinto.scripts.mit.edu/">Nicolas Pinto</a></li>
<li>11 <a class="reference external" href="http://parietal.saclay.inria.fr/Members/virgile-fritsch">Virgile Fritsch</a></li>
<li>7 Lars Buitinck</li>
<li>6 Vincent Michel</li>
<li>5 <a class="reference external" href="http://parietal.saclay.inria.fr/Members/bertrand-thirion">Bertrand Thirion</a></li>
<li>4 Thouis (Ray) Jones</li>
<li>4 Vincent Schut</li>
<li>3 Jan Schlüter</li>
<li>2 Julien Miotte</li>
<li>2 <a class="reference external" href="http://www.lnao.fr/spip.php?rubrique19">Matthieu Perrot</a></li>
<li>2 Yann Malet</li>
<li>2 <a class="reference external" href="http://www.onerussian.com/">Yaroslav Halchenko</a></li>
<li>1 Amit Aides</li>
<li>1 <a class="reference external" href="http://peekaboo-vision.blogspot.com">Andreas Müller</a></li>
<li>1 Feth Arezki</li>
<li>1 Meng Xinfan</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="changes-0-7">
<span id="id17"></span><h1>0.7<a class="headerlink" href="#changes-0-7" title="Permalink to this headline">¶</a></h1>
<p>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.</p>
<div class="section" id="id18">
<h2>Changelog<a class="headerlink" href="#id18" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>Performance improvements for Gaussian Mixture Model sampling [Jan
Schlüter].</li>
<li>Implementation of efficient leave-one-out cross-validated Ridge in
<a class="reference internal" href="modules/generated/sklearn.linear_model.RidgeCV.html#sklearn.linear_model.RidgeCV" title="sklearn.linear_model.RidgeCV"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.RidgeCV</span></tt></a> [<a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>]</li>
<li>Better handling of collinearity and early stopping in
<a class="reference internal" href="modules/generated/sklearn.linear_model.lars_path.html#sklearn.linear_model.lars_path" title="sklearn.linear_model.lars_path"><tt class="xref py py-func docutils literal"><span class="pre">linear_model.lars_path</span></tt></a> [<a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a> and <a class="reference external" href="http://fseoane.net/blog/">Fabian
Pedregosa</a>].</li>
<li>Fixes for liblinear ordering of labels and sign of coefficients
[Dan Yamins, Paolo Losi, <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a> and <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>].</li>
<li>Performance improvements for Nearest Neighbors algorithm in
high-dimensional spaces [<a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>].</li>
<li>Performance improvements for <a class="reference internal" href="modules/generated/sklearn.cluster.KMeans.html#sklearn.cluster.KMeans" title="sklearn.cluster.KMeans"><tt class="xref py py-class docutils literal"><span class="pre">cluster.KMeans</span></tt></a> [<a class="reference external" href="http://gael-varoquaux.info">Gael
Varoquaux</a> and <a class="reference external" href="http://www-etud.iro.umontreal.ca/~bergstrj/">James Bergstra</a>].</li>
<li>Sanity checks for SVM-based classes [<a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a>].</li>
<li>Refactoring of <tt class="xref py py-class docutils literal"><span class="pre">neighbors.NeighborsClassifier</span></tt> and
<a class="reference internal" href="modules/generated/sklearn.neighbors.kneighbors_graph.html#sklearn.neighbors.kneighbors_graph" title="sklearn.neighbors.kneighbors_graph"><tt class="xref py py-func docutils literal"><span class="pre">neighbors.kneighbors_graph</span></tt></a>: added different algorithms for
the k-Nearest Neighbor Search and implemented a more stable
algorithm for finding barycenter weigths. Also added some
developer documentation for this module, see
<a class="reference external" href="https://github.com/scikit-learn/scikit-learn/wiki/Neighbors-working-notes">notes_neighbors</a> for more information [<a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>].</li>
<li>Documentation improvements: Added <tt class="xref py py-class docutils literal"><span class="pre">pca.RandomizedPCA</span></tt> and
<a class="reference internal" href="modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegression" title="sklearn.linear_model.LogisticRegression"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LogisticRegression</span></tt></a> to the class
reference. Also added references of matrices used for clustering
and other fixes [<a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a>, <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>, <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu
Blondel</a>, <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a>, Virgile Fritsch , Emmanuelle
Gouillart]</li>
<li>Binded decision_function in classes that make use of <a class="reference external" href="http://www.csie.ntu.edu.tw/~cjlin/liblinear/">liblinear</a>,
dense and sparse variants, like <a class="reference internal" href="modules/generated/sklearn.svm.LinearSVC.html#sklearn.svm.LinearSVC" title="sklearn.svm.LinearSVC"><tt class="xref py py-class docutils literal"><span class="pre">svm.LinearSVC</span></tt></a> or
<a class="reference internal" href="modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegression" title="sklearn.linear_model.LogisticRegression"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LogisticRegression</span></tt></a> [<a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a>].</li>
<li>Performance and API improvements to
<tt class="xref py py-func docutils literal"><span class="pre">metrics.euclidean_distances</span></tt> and to
<tt class="xref py py-class docutils literal"><span class="pre">pca.RandomizedPCA</span></tt> [<a class="reference external" href="http://www-etud.iro.umontreal.ca/~bergstrj/">James Bergstra</a>].</li>
<li>Fix compilation issues under NetBSD [Kamel Ibn Hassen Derouiche]</li>
<li>Allow input sequences of different lengths in <a class="reference internal" href="modules/generated/sklearn.hmm.GaussianHMM.html#sklearn.hmm.GaussianHMM" title="sklearn.hmm.GaussianHMM"><tt class="xref py py-class docutils literal"><span class="pre">hmm.GaussianHMM</span></tt></a>
[<a class="reference external" href="http://www.ee.columbia.edu/~ronw/">Ron Weiss</a>].</li>
<li>Fix bug in affinity propagation caused by incorrect indexing [Xinfan Meng]</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id19">
<h2>People<a class="headerlink" href="#id19" title="Permalink to this headline">¶</a></h2>
<p>People that made this release possible preceeded by number of commits:</p>
<blockquote>
<div><ul class="simple">
<li>85 <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a></li>
<li>67 <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>20 <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>19 <a class="reference external" href="http://www-etud.iro.umontreal.ca/~bergstrj/">James Bergstra</a></li>
<li>14 Dan Yamins</li>
<li>13 <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>12 <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>4 <a class="reference external" href="http://www.lnao.fr/spip.php?rubrique30">Edouard Duchesnay</a></li>
<li>4 <a class="reference external" href="http://www.ee.columbia.edu/~ronw/">Ron Weiss</a></li>
<li>2 Satrajit Ghosh</li>
<li>2 Vincent Dubourg</li>
<li>1 Emmanuelle Gouillart</li>
<li>1 Kamel Ibn Hassen Derouiche</li>
<li>1 Paolo Losi</li>
<li>1 VirgileFritsch</li>
<li>1 <a class="reference external" href="http://www.onerussian.com/">Yaroslav Halchenko</a></li>
<li>1 Xinfan Meng</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="changes-0-6">
<span id="id20"></span><h1>0.6<a class="headerlink" href="#changes-0-6" title="Permalink to this headline">¶</a></h1>
<p>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.</p>
<div class="section" id="id21">
<h2>Changelog<a class="headerlink" href="#id21" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>New <a class="reference external" href="http://scikit-learn.org/stable/modules/sgd.html">stochastic gradient</a> descent
module by Peter Prettenhofer. The module comes with complete
documentation and examples.</li>
<li>Improved svm module: memory consumption has been reduced by 50%,
heuristic to automatically set class weights, possibility to
assign weights to samples (see
<a class="reference internal" href="auto_examples/svm/plot_weighted_samples.html#example-svm-plot-weighted-samples-py"><em>SVM: Weighted samples</em></a> for an example).</li>
<li>New <a class="reference internal" href="modules/gaussian_process.html#gaussian-process"><em>Gaussian Processes</em></a> module by Vincent Dubourg. This module
also has great documentation and some very neat examples. See
<a class="reference internal" href="auto_examples/gaussian_process/plot_gp_regression.html#example-gaussian-process-plot-gp-regression-py"><em>Gaussian Processes regression: basic introductory example</em></a> or
<a class="reference internal" href="auto_examples/gaussian_process/plot_gp_probabilistic_classification_after_regression.html#example-gaussian-process-plot-gp-probabilistic-classification-after-regression-py"><em>Gaussian Processes classification example: exploiting the probabilistic output</em></a>
for a taste of what can be done.</li>
<li>It is now possible to use liblinear’s Multi-class SVC (option
multi_class in <a class="reference internal" href="modules/generated/sklearn.svm.LinearSVC.html#sklearn.svm.LinearSVC" title="sklearn.svm.LinearSVC"><tt class="xref py py-class docutils literal"><span class="pre">svm.LinearSVC</span></tt></a>)</li>
<li>New features and performance improvements of text feature
extraction.</li>
<li>Improved sparse matrix support, both in main classes
(<a class="reference internal" href="modules/generated/sklearn.grid_search.GridSearchCV.html#sklearn.grid_search.GridSearchCV" title="sklearn.grid_search.GridSearchCV"><tt class="xref py py-class docutils literal"><span class="pre">grid_search.GridSearchCV</span></tt></a>) as in modules
sklearn.svm.sparse and sklearn.linear_model.sparse.</li>
<li>Lots of cool new examples and a new section that uses real-world
datasets was created. These include:
<a class="reference internal" href="auto_examples/applications/face_recognition.html#example-applications-face-recognition-py"><em>Faces recognition example using eigenfaces and SVMs</em></a>,
<a class="reference internal" href="auto_examples/applications/plot_species_distribution_modeling.html#example-applications-plot-species-distribution-modeling-py"><em>Species distribution modeling</em></a>,
<a class="reference internal" href="auto_examples/applications/svm_gui.html#example-applications-svm-gui-py"><em>Libsvm GUI</em></a>,
<a class="reference internal" href="auto_examples/applications/wikipedia_principal_eigenvector.html#example-applications-wikipedia-principal-eigenvector-py"><em>Wikipedia principal eigenvector</em></a> and
others.</li>
<li>Faster <a class="reference internal" href="modules/linear_model.html#least-angle-regression"><em>Least Angle Regression</em></a> algorithm. It is now 2x
faster than the R version on worst case and up to 10x times faster
on some cases.</li>
<li>Faster coordinate descent algorithm. In particular, the full path
version of lasso (<a class="reference internal" href="modules/generated/sklearn.linear_model.lasso_path.html#sklearn.linear_model.lasso_path" title="sklearn.linear_model.lasso_path"><tt class="xref py py-func docutils literal"><span class="pre">linear_model.lasso_path</span></tt></a>) is more than
200x times faster than before.</li>
<li>It is now possible to get probability estimates from a
<a class="reference internal" href="modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegression" title="sklearn.linear_model.LogisticRegression"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LogisticRegression</span></tt></a> model.</li>
<li>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.</li>
<li>Lots of bug fixes and documentation improvements.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id22">
<h2>People<a class="headerlink" href="#id22" title="Permalink to this headline">¶</a></h2>
<p>People that made this release possible preceeded by number of commits:</p>
<blockquote>
<div><ul class="simple">
<li>207 <a class="reference external" href="http://twitter.com/ogrisel">Olivier Grisel</a></li>
<li>167 <a class="reference external" href="http://fseoane.net/blog/">Fabian Pedregosa</a></li>
<li>97 <a class="reference external" href="http://sites.google.com/site/peterprettenhofer/">Peter Prettenhofer</a></li>
<li>68 <a class="reference external" href="http://alexandre.gramfort.net">Alexandre Gramfort</a></li>
<li>59 <a class="reference external" href="http://www.mblondel.org/journal/">Mathieu Blondel</a></li>
<li>55 <a class="reference external" href="http://gael-varoquaux.info">Gael Varoquaux</a></li>
<li>33 Vincent Dubourg</li>
<li>21 <a class="reference external" href="http://www.ee.columbia.edu/~ronw/">Ron Weiss</a></li>
<li>9 Bertrand Thirion</li>
<li>3 <a class="reference external" href="<http://atpassos.posterous.com>">Alexandre Passos</a></li>
<li>3 Anne-Laure Fouque</li>
<li>2 Ronan Amicel</li>
<li>1 <a class="reference external" href="http://osdf.github.com/">Christian Osendorfer</a></li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="changes-0-5">
<span id="id23"></span><h1>0.5<a class="headerlink" href="#changes-0-5" title="Permalink to this headline">¶</a></h1>
<div class="section" id="id24">
<h2>Changelog<a class="headerlink" href="#id24" title="Permalink to this headline">¶</a></h2>
</div>
<div class="section" id="new-classes">
<h2>New classes<a class="headerlink" href="#new-classes" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>Support for sparse matrices in some classifiers of modules
<tt class="docutils literal"><span class="pre">svm</span></tt> and <tt class="docutils literal"><span class="pre">linear_model</span></tt> (see <tt class="xref py py-class docutils literal"><span class="pre">svm.sparse.SVC</span></tt>,
<tt class="xref py py-class docutils literal"><span class="pre">svm.sparse.SVR</span></tt>, <tt class="xref py py-class docutils literal"><span class="pre">svm.sparse.LinearSVC</span></tt>,
<tt class="xref py py-class docutils literal"><span class="pre">linear_model.sparse.Lasso</span></tt>, <tt class="xref py py-class docutils literal"><span class="pre">linear_model.sparse.ElasticNet</span></tt>)</li>
<li>New <a class="reference internal" href="modules/generated/sklearn.pipeline.Pipeline.html#sklearn.pipeline.Pipeline" title="sklearn.pipeline.Pipeline"><tt class="xref py py-class docutils literal"><span class="pre">pipeline.Pipeline</span></tt></a> object to compose different estimators.</li>
<li>Recursive Feature Elimination routines in module
<a class="reference internal" href="modules/feature_selection.html#feature-selection"><em>Feature selection</em></a>.</li>
<li>Addition of various classes capable of cross validation in the
linear_model module (<a class="reference internal" href="modules/generated/sklearn.linear_model.LassoCV.html#sklearn.linear_model.LassoCV" title="sklearn.linear_model.LassoCV"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LassoCV</span></tt></a>, <a class="reference internal" href="modules/generated/sklearn.linear_model.ElasticNetCV.html#sklearn.linear_model.ElasticNetCV" title="sklearn.linear_model.ElasticNetCV"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.ElasticNetCV</span></tt></a>,
etc.).</li>
<li>New, more efficient LARS algorithm implementation. The Lasso
variant of the algorithm is also implemented. See
<a class="reference internal" href="modules/generated/sklearn.linear_model.lars_path.html#sklearn.linear_model.lars_path" title="sklearn.linear_model.lars_path"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.lars_path</span></tt></a>, <a class="reference internal" href="modules/generated/sklearn.linear_model.Lars.html#sklearn.linear_model.Lars" title="sklearn.linear_model.Lars"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.Lars</span></tt></a> and
<a class="reference internal" href="modules/generated/sklearn.linear_model.LassoLars.html#sklearn.linear_model.LassoLars" title="sklearn.linear_model.LassoLars"><tt class="xref py py-class docutils literal"><span class="pre">linear_model.LassoLars</span></tt></a>.</li>
<li>New Hidden Markov Models module (see classes
<a class="reference internal" href="modules/generated/sklearn.hmm.GaussianHMM.html#sklearn.hmm.GaussianHMM" title="sklearn.hmm.GaussianHMM"><tt class="xref py py-class docutils literal"><span class="pre">hmm.GaussianHMM</span></tt></a>, <a class="reference internal" href="modules/generated/sklearn.hmm.MultinomialHMM.html#sklearn.hmm.MultinomialHMM" title="sklearn.hmm.MultinomialHMM"><tt class="xref py py-class docutils literal"><span class="pre">hmm.MultinomialHMM</span></tt></a>,
<a class="reference internal" href="modules/generated/sklearn.hmm.GMMHMM.html#sklearn.hmm.GMMHMM" title="sklearn.hmm.GMMHMM"><tt class="xref py py-class docutils literal"><span class="pre">hmm.GMMHMM</span></tt></a>)</li>
<li>New module feature_extraction (see <a class="reference internal" href="modules/classes.html#feature-extraction-ref"><em>class reference</em></a>)</li>
<li>New FastICA algorithm in module sklearn.fastica</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="documentation">
<h2>Documentation<a class="headerlink" href="#documentation" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>Improved documentation for many modules, now separating
narrative documentation from the class reference. As an example,
see <a class="reference external" href="http://scikit-learn.org/stable/modules/svm.html">documentation for the SVM module</a> and the
complete <a class="reference external" href="http://scikit-learn.org/stable/modules/classes.html">class reference</a>.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="fixes">
<h2>Fixes<a class="headerlink" href="#fixes" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>API changes: adhere variable names to PEP-8, give more
meaningful names.</li>
<li>Fixes for svm module to run on a shared memory context
(multiprocessing).</li>
<li>It is again possible to generate latex (and thus PDF) from the
sphinx docs.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="examples">
<h2>Examples<a class="headerlink" href="#examples" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>new examples using some of the mlcomp datasets:
<a class="reference internal" href="auto_examples/mlcomp_sparse_document_classification.html#example-mlcomp-sparse-document-classification-py"><em>Classification of text documents: using a MLComp dataset</em></a>,
<a class="reference internal" href="auto_examples/document_classification_20newsgroups.html#example-document-classification-20newsgroups-py"><em>Classification of text documents using sparse features</em></a></li>
<li>Many more examaples. <a class="reference external" href="http://scikit-learn.org/stable/auto_examples/index.html">See here</a>
the full list of examples.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="external-dependencies">
<h2>External dependencies<a class="headerlink" href="#external-dependencies" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>Joblib is now a dependencie of this package, although it is
shipped with (sklearn.externals.joblib).</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="removed-modules">
<h2>Removed modules<a class="headerlink" href="#removed-modules" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>Module ann (Artificial Neural Networks) has been removed from
the distribution. Users wanting this sort of algorithms should
take a look into pybrain.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="misc">
<h2>Misc<a class="headerlink" href="#misc" title="Permalink to this headline">¶</a></h2>
<blockquote>
<div><ul class="simple">
<li>New sphinx theme for the web page.</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="authors">
<h2>Authors<a class="headerlink" href="#authors" title="Permalink to this headline">¶</a></h2>
<p>The following is a list of authors for this release, preceeded by
number of commits:</p>
<blockquote>
<div><ul class="simple">
<li>262 Fabian Pedregosa</li>
<li>240 Gael Varoquaux</li>
<li>149 Alexandre Gramfort</li>
<li>116 Olivier Grisel</li>
<li>40 Vincent Michel</li>
<li>38 Ron Weiss</li>
<li>23 Matthieu Perrot</li>
<li>10 Bertrand Thirion</li>
<li>7 Yaroslav Halchenko</li>
<li>9 VirgileFritsch</li>
<li>6 Edouard Duchesnay</li>
<li>4 Mathieu Blondel</li>
<li>1 Ariel Rokem</li>
<li>1 Matthieu Brucher</li>
</ul>
</div></blockquote>
</div>
</div>
<div class="section" id="id25">
<h1>0.4<a class="headerlink" href="#id25" title="Permalink to this headline">¶</a></h1>
<div class="section" id="id26">
<h2>Changelog<a class="headerlink" href="#id26" title="Permalink to this headline">¶</a></h2>
<p>Major changes in this release include:</p>
<blockquote>
<div><ul class="simple">
<li>Coordinate Descent algorithm (Lasso, ElasticNet) refactoring &
speed improvements (roughly 100x times faster).</li>
<li>Coordinate Descent Refactoring (and bug fixing) for consistency
with R’s package GLMNET.</li>
<li>New metrics module.</li>
<li>New GMM module contributed by Ron Weiss.</li>
<li>Implementation of the LARS algorithm (without Lasso variant for now).</li>
<li>feature_selection module redesign.</li>
<li>Migration to GIT as content management system.</li>
<li>Removal of obsolete attrselect module.</li>
<li>Rename of private compiled extensions (aded underscore).</li>
<li>Removal of legacy unmaintained code.</li>
<li>Documentation improvements (both docstring and rst).</li>
<li>Improvement of the build system to (optionally) link with MKL.
Also, provide a lite BLAS implementation in case no system-wide BLAS is
found.</li>
<li>Lots of new examples.</li>
<li>Many, many bug fixes ...</li>
</ul>
</div></blockquote>
</div>
<div class="section" id="id27">
<h2>Authors<a class="headerlink" href="#id27" title="Permalink to this headline">¶</a></h2>
<p>The committer list for this release is the following (preceded by number
of commits):</p>
<blockquote>
<div><ul class="simple">
<li>143 Fabian Pedregosa</li>
<li>35 Alexandre Gramfort</li>
<li>34 Olivier Grisel</li>
<li>11 Gael Varoquaux</li>
<li>5 Yaroslav Halchenko</li>
<li>2 Vincent Michel</li>
<li>1 Chris Filo Gorgolewski</li>
</ul>
</div></blockquote>
</div>
</div>
</div>
</div>
</div>
<div class="clearer"></div>
</div>
</div>
<div class="footer">
© 2010–2011, scikit-learn developers (BSD License).
Created using <a href="http://sphinx.pocoo.org/">Sphinx</a> 1.1.3. Design by <a href="http://desgrana.es">Desgrana</a>.
<span style="padding-left: 5ex;">
<a href="_sources/whats_new.txt"
rel="nofollow">Show this page source</a>
</span>
</div>
<div class="rel rellarge">
<div class="buttonPrevious">
<a href="support.html">
Previous
</a>
</div>
<div class="buttonNext">
<a href="presentations.html">
Next
</a>
</div>
</div>
<script type="text/javascript">
$("div.buttonNext, div.buttonPrevious").hover(
function () {
$(this).css('background-color', '#FF9C34');
},
function () {
$(this).css('background-color', '#A7D6E2');
}
);
</script>
</body>
</html>