Revisiting Non-Acyclic GFlowNets in Discrete Environments

P. 44887–44910.

Morozov N., Maximov I., Tiapkin D., Samsonov S.

Generative Flow Networks (GFlowNets) are a family of generative models that learn to sample objects from a given probability distribution, potentially known up to a normalizing constant. Instead of working in the object space, GFlowNets proceed by sampling trajectories in an appropriately constructed directed acyclic graph environment, greatly relying on the acyclicity of the graph. In our paper, we revisit the theory that relaxes the acyclicity assumption and present a simpler theoretical framework for non-acyclic GFlowNets in discrete environments. Moreover, we provide various novel theoretical insights related to training with fixed backward policies, the nature of flow functions, and connections between entropy-regularized RL and non-acyclic GFlowNets, which naturally generalize the respective concepts and theoretical results from the acyclic setting. In addition, we experimentally re-examine the concept of loss stability in nonacyclic GFlowNet training, as well as validate our own theoretical findings

Keywords: generative models, Machine Learning, Generative Flow Networks

Publication based on the results of: Development of theoretical foundations and methods of generative artificial intelligence and their application to heterogeneous domain area (2025)

In book

Volume 267: International Conference on Machine Learning, 13-19 July 2025, Vancouver Convention Center, Vancouver, Canada

Vol. 267. , [б.и.], 2025.