Tpot

In TPOT 0.4, we've made some major changes to the internals of TPOT and added some convenience functions. We've summarized the changes below.

<ul>
<li>Added new sklearn models and preprocessors
<ul>
<li>AdaBoostClassifier</li>
<li>BernoulliNB</li>
<li>ExtraTreesClassifier</li>
<li>GaussianNB</li>
<li>MultinomialNB</li>
<li>LinearSVC</li>
<li>PassiveAggressiveClassifier</li>
<li>GradientBoostingClassifier</li>
<li>RBFSampler</li>
<li>FastICA</li>
<li>FeatureAgglomeration</li>
<li>Nystroem</li>
</ul></li>
<li>Added operator that inserts virtual features for the count of features with values of zero</li>
<li>Reworked parameterization of TPOT operators
<ul>
<li>Reduced parameter search space with information from a scikit-learn benchmark</li>
<li>TPOT no longer generates arbitrary parameter values, but uses a fixed parameter set instead</li>
</ul></li>
<li>Removed XGBoost as a dependency
<ul>
<li>Too many users were having install issues with XGBoost</li>
<li>Replaced with scikit-learn's GradientBoostingClassifier</li>
</ul></li>
<li>Improved descriptiveness of TPOT command line parameter documentation</li>
<li>Removed min/max/avg details during fit() when verbosity &gt; 1
<ul>
<li>Replaced with tqdm progress bar</li>
<li>Added tqdm as a dependency</li>
</ul></li>
<li>Added <code>fit_predict()</code> convenience function</li>
<li>Added <code>get_params()</code> function so TPOT can operate in scikit-learn's <code>cross_val_score</code> & related functions</li>
</ul>

Zenodo requires me to make a new release to assign a DOI, so here's that release. This is not a full release.

This is the version of TPOT that was used in the GECCO 2016 paper, "Evaluation of a Tree-based Pipeline Optimization Tool for Automating Data Science."

- TPOT now has the ability to export the optimized pipelines to sklearn code. See the [documentation](http://rhiever.github.io/tpot/examples/Using_TPOT_via_code/) for more information.
- Logistic regression, SVM, and k-nearest neighbors classifiers were added as pipeline operators. Previously, TPOT only included decision tree and random forest classifiers.
- TPOT can now use arbitrary scoring functions for the optimization process. See the [scoring function documentation](http://rhiever.github.io/tpot/examples/Custom_Scoring_Functions/) for more information.