GFlowNet Foundations

By Yoshua Bengio, Tristan Deleu, J. Edward Hu, Salem Lahlou, Mo Tiwari, Emmanuel Bengio. 2021

363 citations recorded by the source corpus.

Generative Flow Networks (GFlowNets) have been introduced as a method to sample a diverse set of candidates in an active learning context, with a training objective that makes them approximately sample in proportion to a given reward function. In this paper, we show a number of additional theoretical properties of GFlowNets. They can be used to estimate joint probability distributions and the corresponding marginal distributions where some variables are unspecified and, of particular interest, can represent distributions over composite objects like sets and graphs. GFlowNets amortize the work typically done by computationally expensive MCMC methods in a single but trained generative pass. They could also be used to estimate partition functions and free energies, conditional probabilities of supersets (supergraphs) given a subset (subgraph), as well as marginal distributions over all supersets (supergraphs) of a given set (graph). We introduce variations enabling the estimation of entropy and mutual information, sampling from a Pareto frontier, connections to reward-maximizing policies, and extensions to stochastic environments, continuous actions and modular energy functions.

Subfields: Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Reinforcement Learning

Citations in the corpus

Bayesian Structure Learning with Generative Flow Networks (2022, methodology): “Recall that the detailed-balance condition (Bengio et al., 2021b, Def.”
MolGenSurvey: A Systematic Survey in Machine Learning Models for Molecule Design (2022, background): “Speciﬁcally, [108], [109], [110], [111] leverage policy gradient and domain-speciﬁc reward function to optimize the graph policy neural network; [103] leverages deep Q-network to conduct value learning, where a deep Q-network is trained to predict the future reward and during each step, the step…”