Metadata-Version: 1.1
Name: nolearn
Version: 0.2
Summary: Miscellaneous utilities for machine learning.
Home-page: https://github.com/dnouri/nolearn
Author: Daniel Nouri
Author-email: daniel.nouri@gmail.com
License: MIT
Description: This package contains a number of utility modules that are helpful
        with machine learning tasks.  Most of the modules work together with
        `scikit-learn <http://scikit-learn.org/>`_, others are more generally
        useful.
        
        View the `documentation here <http://packages.python.org/nolearn/>`_.
        
        
        Change History
        ==============
        
        0.2 - 2013-03-03
        ----------------
        
        - Add parameters `learn_rate_decays` and `learn_rate_minimums`, which
          allow for decreasing the learning after each epoch of fine-tuning.
        
        - Allow `-1` as the value of the input and output layers of the neural
          network.  The shapes of `X` and `y` will then be used to determine
          those.
        
        - Add support for processing sparse input data matrices.
        
        - Improve miserable performance of `DBN.predict_proba`.
        
        0.2b1 - 2012-12-30
        ------------------
        
        - Added a scikit-learn estimator based on George Dahl's `gdbn` in
          `nolearn.dbn`.
        
Platform: UNKNOWN
Classifier: Development Status :: 4 - Beta
