End-to-End Learning under Endogenous Uncertainty

Rares C Cristian, Pavithra Harsha, Georgia Perakis, Brian Quanz

Keywords: End-to-end learning, contextual stochastic optimization

Abstract:

How can we effectively learn to make decisions when there are no ground-truth counterfactual observations? We propose an end-to-end learning approach to the contextual stochastic optimization problem under decision-dependent uncertainty. We propose both exact methods and efficient sampling-based methods to implement our approach. We also introduce a new class of two-stage stochastic optimization problems to the end-to-end learning framework. Here, the first stage is an information-gathering problem to decide which random variable to ``poll'' and gain information about before making a second-stage decision based off of it. We provide theoretical analysis showing (1) that optimally minimizing our proposed objective produces optimal decisions and (2) generalization bounds between in-sample and out-of-sample cost.
We computationally test the proposed approach on multi-item assortment problems where demand is affected by cross-item complementary and supplementary effects. Overall, our method outperforms other benchmarks by more than 15\% and performs best in high noise, across any cost configuration, and when given sufficient data.
We also introduce an experiment for the information-gathering problem on a real-world electricity generation problem. We show our method proposes decisions with more than 7\% lower cost than other decision-making methods.

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.

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: 7879