Search papers, labs, and topics across Lattice.
This paper introduces a GFlowNet-based framework for learning shortest paths in graphs, even when non-acyclic. It proves that minimizing total flow in a GFlowNet forces forward and backward policies to converge on shortest paths. Experiments demonstrate the method's effectiveness in permutation environments and Rubik's Cube solving, achieving competitive solution lengths with smaller search budgets compared to specialized ML methods.
GFlowNets can provably learn shortest paths in arbitrary graphs by minimizing total flow, outperforming specialized methods on Rubik's Cube with less search.
In this paper, we present a novel learning framework for finding shortest paths in graphs utilizing Generative Flow Networks (GFlowNets). First, we examine theoretical properties of GFlowNets in non-acyclic environments in relation to shortest paths. We prove that, if the total flow is minimized, forward and backward policies traverse the environment graph exclusively along shortest paths between the initial and terminal states. Building on this result, we show that the pathfinding problem in an arbitrary graph can be solved by training a non-acyclic GFlowNet with flow regularization. We experimentally demonstrate the performance of our method in pathfinding in permutation environments and in solving Rubik's Cubes. For the latter problem, our approach shows competitive results with state-of-the-art machine learning approaches designed specifically for this task in terms of the solution length, while requiring smaller search budget at test-time.
