This repository contains the implementation for the paper Information Theoretic StructuredGenerative Modeling,

Specially thanks for the open-source codes shared by sagelywizard/pytorch-mdn and PyTorch-GAN

Main Requirements

• Pytorch
• A GPU Machine

Usage

The main experiments in the paper are put in the notebook format.

Each file can be run independently

• GAN Example
• MINE Example
• Density Estimation
• Conditonal Estimation

To run ood_visualization.ipynb, please download the pretrained model in the ./model/ folder. (google drive link later)

The other baselines can be run by calling

python MDN.py
python CGAN.py

GMM_VBGMM_CE.py provides codes for producing conditional CE for any mixture models obtained from scipy.

Simply calling

compute_conditionalCE(joint, gm_joint)

in python to obtain the value, where joint should be bs*K and gm_joint is the class obtained from scipy.