GPU-based python implementation of
- Feed-forward Neural Nets
- Restricted Boltzmann Machines
- Deep Belief Nets
- Autoencoders
- Deep Boltzmann Machines
- Convolutional Nets
Code
Built on top of the cudamat library by Vlad Mnih and cuda-convnet library by Alex Krizhevsky.
Running the examples
- Get the MNIST data set from here .
- cd to the example directory (e.g. examples/ff)
- Edit the paths in the train.pbtxt file appropriately.
- Execute runall.sh
Multimodal Deep Belief Nets and Deep Boltzmann Machines on Flickr [ICML 2012, NIPS 2012]
Model files and trainers on the project page
Dropout results [paper]