Self-Evaluation for Job-Shop Scheduling
Imanol Echeverria, Maialen Murua, Roberto Santana
TL;DR
This work targets NP-hard combinatorial optimization, focusing on Job-Shop Scheduling. It introduces SEVAL, a subsequence-based self-evaluation framework that combines a heterogeneous graph neural network and a Transformer policy with a Transformer-based self-evaluation module to score candidate action subsets. By recasting COPs as MDPs over subsequences and training offline with solver-provided data, SEVAL achieves state-of-the-art makespan performance on public JSSP benchmarks, including strong generalization to distribution-shifted instances, while remaining computationally efficient. The approach advances practical decision-making in scheduling by reducing error propagation in sequential decisions and enabling robust, high-quality solution generation for large-scale problems.
Abstract
Combinatorial optimization problems, such as scheduling and route planning, are crucial in various industries but are computationally intractable due to their NP-hard nature. Neural Combinatorial Optimization methods leverage machine learning to address these challenges but often depend on sequential decision-making, which is prone to error accumulation as small mistakes propagate throughout the process. Inspired by self-evaluation techniques in Large Language Models, we propose a novel framework that generates and evaluates subsets of assignments, moving beyond traditional stepwise approaches. Applied to the Job-Shop Scheduling Problem, our method integrates a heterogeneous graph neural network with a Transformer to build a policy model and a self-evaluation function. Experimental validation on challenging, well-known benchmarks demonstrates the effectiveness of our approach, surpassing state-of-the-art methods.