Ant Colony Sampling with GFlowNets for Combinatorial Optimization

TL;DR

This paper introduces GFACS, a new meta-heuristic combining GFlowNets and Ant Colony Optimization to efficiently explore and optimize complex combinatorial solution spaces, showing promising experimental results.

Contribution

The paper proposes GFACS, a novel hierarchical method that integrates amortized inference with parallel stochastic search for combinatorial optimization.

Findings

GFACS effectively approximates multi-modal solution distributions.

The method achieves near-optimal solutions across diverse problems.

Experimental results demonstrate competitive performance.

Abstract

We present the Generative Flow Ant Colony Sampler (GFACS), a novel meta-heuristic method that hierarchically combines amortized inference and parallel stochastic search. Our method first leverages Generative Flow Networks (GFlowNets) to amortize a \emph{multi-modal} prior distribution over combinatorial solution space that encompasses both high-reward and diversified solutions. This prior is iteratively updated via parallel stochastic search in the spirit of Ant Colony Optimization (ACO), leading to the posterior distribution that generates near-optimal solutions. Extensive experiments across seven combinatorial optimization problems demonstrate GFACS's promising performances.