/dgm

Deep Generative Model (Torch)

Primary LanguageLuaMIT LicenseMIT

Deep Generative Model

This contains the Torch Implementation of a Deep Generative Model trained using Stochastic Backpropagation. The model is trained on the GPU with binarized MNIST.

Requirements

Torch installed with cunn

Some miscellaneous lua packages installed from: https://raw.githubusercontent.com/rahulk90/helper-files/master/install_torch_deps.sh

Learning the Model

Run th train.lua

Uses display (https://github.com/szym/display) to show samples and reconstruction to a browser display

References

Auto-Encoding Variational Bayes http://arxiv.org/abs/1312.6114

Stochastic Backpropagation and Approximate Inference in Deep Generative Models http://arxiv.org/abs/1401.4082

See also: https://github.com/y0ast/VAE-Torch