Neural Variational Inference Document Modeling

This code implements the Neural Inference Document Modeling (NVDM) in the paper "Neural Variational Inference for Text Processing".

To compute the perplexity for test document, if you use the latest ClassNLLCriterion, its parameter sizeAverage should be set true.

I cannot get good results yet. Another Tensorflow implementation is at "(https://github.com/carpedm20/variational-text-tensorflow)".