Bifurcated Generative Flow Networks

TL;DR

Bifurcated GFlowNets introduce a new architecture that improves data efficiency and scalability in probabilistic sampling tasks, maintaining convergence guarantees and outperforming existing methods.

Contribution

The paper proposes Bifurcated GFlowNets, a novel architecture that factorizes flows to enhance learning efficiency and scalability in GFlowNets.

Findings

BN significantly improves learning efficiency.

BN outperforms strong baselines on benchmarks.

BN maintains convergence guarantees.

Abstract

Generative Flow Networks (GFlowNets), a new family of probabilistic samplers, have recently emerged as a promising framework for learning stochastic policies that generate high-quality and diverse objects proportionally to their rewards. However, existing GFlowNets often suffer from low data efficiency due to the direct parameterization of edge flows or reliance on backward policies that may struggle to scale up to large action spaces. In this paper, we introduce Bifurcated GFlowNets (BN), a novel approach that employs a bifurcated architecture to factorize the flows into separate representations for state flows and edge-based flow allocation. This factorization enables BN to learn more efficiently from data and better handle large-scale problems while maintaining the convergence guarantee. Through extensive experiments on standard evaluation benchmarks, we demonstrate that BN significantly improves learning efficiency and effectiveness compared to strong baselines.