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
More documentation is available in the form of commented examples scripts and ipython notebooks in the "pylearn2/scripts/tutorials" directory.
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-users Google group for important updates. Please write to this list for general inquiries and support questions.
- Subscribe to the pylearn-dev Google group for important development updates. Please write to this list if you find any bug or want to contribute to the project.
- 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.
- 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.
License and Citations
Pylearn2 is released under the 3-claused BSD license, so it may be used for commercial purposes. The license does not require anyone to cite Pylearn2, but if you use Pylearn2 in published research work we encourage you to cite this article: