Operations Research 2025
Abstract Submission

2312. Interpretable Surrogates for Optimization

Invited abstract in session WC-12: Explainability and Interpretability in Optimization, stream Artificial Intelligence, Machine Learning and Optimization.

Wednesday, 13:30-15:00
Room: H10

Authors (first author is the speaker)

1. Sebastian Merten
Business Decisions and Data Science, University of Passau
2. Marc Goerigk
Business Decisions and Data Science, University of Passau
3. Michael Hartisch
Data Science, Friedrich-Alexander-Universität Erlangen-Nürnberg
4. Kartikey Sharma
AISST, Zuse Institute Berlin

Abstract

An important factor in the practical implementation of optimization models is their acceptance by the intended users. This is influenced by various factors, including the interpretability of the solution process. A recently introduced framework for inherently interpretable optimization models proposes surrogates (e.g. decision trees) of the optimization process. These surrogates represent inherently interpretable rules for mapping problem instances to solutions of the underlying optimization model. In contrast to the use of conventional black-box solution methods, the application of these surrogates thus offers an interpretable solution approach.

Building on this work, we investigate how we can generalize this idea to further increase interpretability while concurrently giving more freedom to the decision maker. We introduce surrogates which do not map to a concrete solution, but to a solution set instead, which is characterized by certain features. Furthermore, we address the question of how to generate surrogates that are better protected against perturbations. We use the concept of robust optimization to generate decision trees that perform well even in the worst case. For both approaches, exact methods as well as heuristics are presented and experimental results are shown. In particular, the relationship between interpretability and performance is discussed.

Keywords

Status: accepted

Back to the list of papers