Optimizing Discrete Spaces via Expensive Evaluations: A Learning to Search Framework

TL;DR

This paper introduces L2S-DISCO, a learning-to-search framework for optimizing expensive black-box functions over discrete spaces, demonstrating improved efficiency over existing methods through empirical evaluation.

Contribution

The paper presents a novel learning-to-search approach, L2S-DISCO, that adaptively guides search procedures for discrete optimization with expensive evaluations.

Findings

L2S-DISCO outperforms state-of-the-art algorithms on real-world benchmarks.

The framework effectively learns control strategies to improve search efficiency.

Empirical results validate the approach's superiority in complex discrete optimization tasks.

Abstract

We consider the problem of optimizing expensive black-box functions over discrete spaces (e.g., sets, sequences, graphs). The key challenge is to select a sequence of combinatorial structures to evaluate, in order to identify high-performing structures as quickly as possible. Our main contribution is to introduce and evaluate a new learning-to-search framework for this problem called L2S-DISCO. The key insight is to employ search procedures guided by control knowledge at each step to select the next structure and to improve the control knowledge as new function evaluations are observed. We provide a concrete instantiation of L2S-DISCO for local search procedure and empirically evaluate it on diverse real-world benchmarks. Results show the efficacy of L2S-DISCO over state-of-the-art algorithms in solving complex optimization problems.