Decision-Focused Learning: Foundations, State of the Art, Benchmark and Future Opportunities

TL;DR

This paper reviews decision-focused learning (DFL), an innovative approach combining machine learning and optimization to improve decision-making under uncertainty, highlighting techniques, empirical results, and future research directions.

Contribution

It provides the first comprehensive survey and benchmark of DFL methods, analyzing their strengths, limitations, and potential for advancing decision-making applications.

Findings

Empirical evaluation of eleven DFL methods across seven problems.

Gradient-based and gradient-free techniques have complementary strengths.

Future research should focus on scalability and real-world applications.

Abstract

Decision-focused learning (DFL) is an emerging paradigm that integrates machine learning (ML) and constrained optimization to enhance decision quality by training ML models in an end-to-end system. This approach shows significant potential to revolutionize combinatorial decision-making in real-world applications that operate under uncertainty, where estimating unknown parameters within decision models is a major challenge. This paper presents a comprehensive review of DFL, providing an in-depth analysis of both gradient-based and gradient-free techniques used to combine ML and constrained optimization. It evaluates the strengths and limitations of these techniques and includes an extensive empirical evaluation of eleven methods across seven problems. The survey also offers insights into recent advancements and future research directions in DFL. Code and benchmark: https://github.com/PredOpt/predopt-benchmarks