Generative Flow Networks for Discrete Probabilistic Modeling

Dinghuai Zhang, Nikolay Malkin, Zhen Liu, Alexandra Volokhova, Aaron Courville, Yoshua Bengio
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:26412-26428, 2022.

Abstract

We present energy-based generative flow networks (EB-GFN), a novel probabilistic modeling algorithm for high-dimensional discrete data. Building upon the theory of generative flow networks (GFlowNets), we model the generation process by a stochastic data construction policy and thus amortize expensive MCMC exploration into a fixed number of actions sampled from a GFlowNet. We show how GFlowNets can approximately perform large-block Gibbs sampling to mix between modes. We propose a framework to jointly train a GFlowNet with an energy function, so that the GFlowNet learns to sample from the energy distribution, while the energy learns with an approximate MLE objective with negative samples from the GFlowNet. We demonstrate EB-GFN’s effectiveness on various probabilistic modeling tasks. Code is publicly available at https://github.com/zdhNarsil/EB_GFN.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-zhang22v, title = {Generative Flow Networks for Discrete Probabilistic Modeling}, author = {Zhang, Dinghuai and Malkin, Nikolay and Liu, Zhen and Volokhova, Alexandra and Courville, Aaron and Bengio, Yoshua}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {26412--26428}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/zhang22v/zhang22v.pdf}, url = {https://proceedings.mlr.press/v162/zhang22v.html}, abstract = {We present energy-based generative flow networks (EB-GFN), a novel probabilistic modeling algorithm for high-dimensional discrete data. Building upon the theory of generative flow networks (GFlowNets), we model the generation process by a stochastic data construction policy and thus amortize expensive MCMC exploration into a fixed number of actions sampled from a GFlowNet. We show how GFlowNets can approximately perform large-block Gibbs sampling to mix between modes. We propose a framework to jointly train a GFlowNet with an energy function, so that the GFlowNet learns to sample from the energy distribution, while the energy learns with an approximate MLE objective with negative samples from the GFlowNet. We demonstrate EB-GFN’s effectiveness on various probabilistic modeling tasks. Code is publicly available at https://github.com/zdhNarsil/EB_GFN.} }
Endnote
%0 Conference Paper %T Generative Flow Networks for Discrete Probabilistic Modeling %A Dinghuai Zhang %A Nikolay Malkin %A Zhen Liu %A Alexandra Volokhova %A Aaron Courville %A Yoshua Bengio %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-zhang22v %I PMLR %P 26412--26428 %U https://proceedings.mlr.press/v162/zhang22v.html %V 162 %X We present energy-based generative flow networks (EB-GFN), a novel probabilistic modeling algorithm for high-dimensional discrete data. Building upon the theory of generative flow networks (GFlowNets), we model the generation process by a stochastic data construction policy and thus amortize expensive MCMC exploration into a fixed number of actions sampled from a GFlowNet. We show how GFlowNets can approximately perform large-block Gibbs sampling to mix between modes. We propose a framework to jointly train a GFlowNet with an energy function, so that the GFlowNet learns to sample from the energy distribution, while the energy learns with an approximate MLE objective with negative samples from the GFlowNet. We demonstrate EB-GFN’s effectiveness on various probabilistic modeling tasks. Code is publicly available at https://github.com/zdhNarsil/EB_GFN.
APA
Zhang, D., Malkin, N., Liu, Z., Volokhova, A., Courville, A. & Bengio, Y.. (2022). Generative Flow Networks for Discrete Probabilistic Modeling. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:26412-26428 Available from https://proceedings.mlr.press/v162/zhang22v.html.

Related Material