Pylearn2: A machine learning research library

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

Pylearn2 has online documentation. If you want to build a local copy of the documentation, run

python ./doc/scripts/docgen.py

More documentation is available in the form of ipython notebooks in the "tutorials" directory, and some example scripts in "scripts/train_example."

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

• Installation instructions are available here.
• 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.
• Read through the documentation and examples mentioned above.
• 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. In fact, you can probably use an existing TrainingAlgorithm.

Highlights

• Pylearn2 was used to set the state of the art on MNIST, CIFAR-10, CIFAR-100, and SVHN. See pylearn2.models.maxout or pylearn2/scripts/papers/maxout
• Pylearn2 provides a wrapper around Alex Krizhevsky's extremely efficient GPU convolutional network library. This wrapper lets you use Theano's symbolic differentiation and other capabilities with minimal overhead. See pylearn2.sandbox.cuda_convnet.