GPU-based python implementation of
- Feed-forward Neural Nets
- Restricted Boltzmann Machines
- Deep Belief Nets
- Autoencoders
- Deep Boltzmann Machines
- Convolutional Neural Nets
Built on top of the cudamat library by Vlad Mnih and cuda-convnet library by Alex Krizhevsky.
Instructions:
- Install Python Tools for Visual Studio
- Install CUDA (tested on CUDA 6.0).
- Compile cudamat
- From the cudamat directory run: "nmake -f Makefile.win"
- Compile eigen-mat
- From the eigenmat directory run: "nmake -f Makefile.win". The automated process will download Eigen3, Powershell 3.0 and .Net 4.5 if they are not available (those 2 are needed to automate the rest).
- Open the solution and enjoy.
Pull requests are welcome.