Pylearn2 is a library designed to make machine learning research easy.

The Pylearn2 documentation. If you want to build a local copy of the documentation, run python ./doc/scripts/docgen.py

Pylearn2 was initially developed by David Warde-Farley, Pascal Lamblin, Ian Goodfellow and others during the winter 2011 offering of IFT6266, and is now developed by the LISA lab.

Quick start and basic design rules ------------------

• Subscribe to the pylearn-dev Google group for important updates. Please write to this list for troubleshooting help or any feedback you have about the library, even if you're not a Pylearn2 developer.
• There are examples. They cover how to create a dataset and how to train and how to inspect the model.
• Pylearn2 should not force users to commit to the whole library. If someone just wants to implement a Model, they should be able to do that and not need to implement a TrainingAlgorithm. Try not to write library features that force users to buy into the whole library.
• When writing reference implementations to go in the library, maximize code re-usability by decomposing your algorithm into a TrainingAlgorithm that trains a Model on a Dataset. It will probably do this by minimizing a Cost.