/GFlowNet-EM

Code for GFlowNet-EM, a novel algorithm for fitting latent variable models with compositional latents and an intractable true posterior.

Primary LanguageJupyter NotebookMIT LicenseMIT

Source code for experiments in "GFlowNet-EM for Learning Compositional Latent Variable Models"

This repository contains code for our paper

GFlowNet-EM for Learning Compositional Latent Variable Models
Edward J. Hu*, Nikolay Malkin*, Moksh Jain, Katie Everett, Alexandros Graikos, Yoshua Bengio
Paper: https://arxiv.org/abs/2302.06576

We provide code for the three experiment domains studied:

  • Illustrative experiment with Gaussian mixtures (toy_gaussians),
  • Grammar induction (grammars),
  • Discrete variational autoencoders (discrete_vae).

Please visit the individual directories for instructions on reproducing the results.

Citation

@inproceedings{hu2023gflownetem,
title={{GFlowNet-EM} for Learning Compositional Latent Variable Models},
author={Hu, Edward J and Malkin, Nikolay and Jain, Moksh and Everett, Katie and Graikos, Alexandros and Bengio, Yoshua},
booktitle={International Conference on Machine Learning},
year={2023},
url={https://arxiv.org/abs/2302.06576}
}

Please contact us or post an issue if you have any questions.