Optimizing Backward Policies in GFlowNets via Trajectory Likelihood Maximization

TL;DR

This paper introduces a backward policy optimization method for GFlowNets that directly maximizes trajectory likelihood, leading to faster convergence and improved mode discovery in complex environments.

Contribution

It proposes a novel backward policy optimization algorithm based on trajectory likelihood maximization, extending GFlowNet training beyond fixed backward policies.

Findings

Faster convergence in complex environments

Enhanced mode discovery capabilities

Effective integration with existing RL and GFlowNet algorithms

Abstract

Generative Flow Networks (GFlowNets) are a family of generative models that learn to sample objects with probabilities proportional to a given reward function. The key concept behind GFlowNets is the use of two stochastic policies: a forward policy, which incrementally constructs compositional objects, and a backward policy, which sequentially deconstructs them. Recent results show a close relationship between GFlowNet training and entropy-regularized reinforcement learning (RL) problems with a particular reward design. However, this connection applies only in the setting of a fixed backward policy, which might be a significant limitation. As a remedy to this problem, we introduce a simple backward policy optimization algorithm that involves direct maximization of the value function in an entropy-regularized Markov Decision Process (MDP) over intermediate rewards. We provide an extensive experimental evaluation of the proposed approach across various benchmarks in combination with both RL and GFlowNet algorithms and demonstrate its faster convergence and mode discovery in complex environments.