Black-Box Combinatorial Optimization with Order-Invariant Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Goudet, Olivier, Suire, Quentin, Goëffon, Adrien, Saubion, Frédéric, Lamprier, Sylvain
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910004739571712
author Goudet, Olivier
Suire, Quentin
Goëffon, Adrien
Saubion, Frédéric
Lamprier, Sylvain
author_facet Goudet, Olivier
Suire, Quentin
Goëffon, Adrien
Saubion, Frédéric
Lamprier, Sylvain
contents We introduce an order-invariant reinforcement learning framework for black-box combinatorial optimization. Classical estimation-of-distribution algorithms (EDAs) often rely on learning explicit variable dependency graphs, which can be costly and fail to capture complex interactions efficiently. In contrast, we parameterize a multivariate autoregressive generative model trained without a fixed variable ordering. By sampling random generation orders during training, a form of information-preserving dropout, the model is encouraged to be invariant to variable order, promoting search-space diversity, and shaping the model to focus on the most relevant variable dependencies, improving sample efficiency. We adapt Group Relative Policy Optimization (GRPO) to this setting, providing stable policy-gradient updates from scale-invariant advantages. Across a wide range of benchmark algorithms and problem instances of varying sizes, our method frequently achieves the best performance and consistently avoids catastrophic failures.
format Preprint
id arxiv_https___arxiv_org_abs_2510_01824
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Black-Box Combinatorial Optimization with Order-Invariant Reinforcement Learning
Goudet, Olivier
Suire, Quentin
Goëffon, Adrien
Saubion, Frédéric
Lamprier, Sylvain
Machine Learning
We introduce an order-invariant reinforcement learning framework for black-box combinatorial optimization. Classical estimation-of-distribution algorithms (EDAs) often rely on learning explicit variable dependency graphs, which can be costly and fail to capture complex interactions efficiently. In contrast, we parameterize a multivariate autoregressive generative model trained without a fixed variable ordering. By sampling random generation orders during training, a form of information-preserving dropout, the model is encouraged to be invariant to variable order, promoting search-space diversity, and shaping the model to focus on the most relevant variable dependencies, improving sample efficiency. We adapt Group Relative Policy Optimization (GRPO) to this setting, providing stable policy-gradient updates from scale-invariant advantages. Across a wide range of benchmark algorithms and problem instances of varying sizes, our method frequently achieves the best performance and consistently avoids catastrophic failures.
title Black-Box Combinatorial Optimization with Order-Invariant Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2510.01824