Explainable Sequential Optimization
Keywords: Explainability, Sequential Optimization
Abstract: We propose formulating stochastic model predictive control into a coalition game to use Shapley values for feature attribution. Such analysis is crucial for transparency and achieving optimal outcomes in high-stake applications such as portfolio optimization and autonomous driving. We categorize Shapley values estimation methods into three families: those based on weighted linear regression, sampling permutations, and multilinear extension. We survey, benchmark, and provide valuable insight into these methods, previously not attempted in this context. Our experiments show that halved Owen sampling from multilinear extension and KernelShap-Paired from weighted linear regression, both utilizing antithetic sampling, perform best.
Primary Area: optimization
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 11520
Loading