Conformal Contextual Robust Optimization

TL;DR

This paper introduces Conformal-Predict-Then-Optimize (CPO), a framework that combines conformal prediction with decision optimization, providing robust, distribution-free uncertainty regions and visual explanations for decision-making in safety-critical, high-dimensional problems.

Contribution

The paper presents a novel CPO framework that integrates conformal prediction with optimization and visualization, enhancing robustness and interpretability in data-driven decision-making.

Findings

CPO provides distribution-free coverage guarantees.

CPO achieves robust decision-making in high-dimensional settings.

Visual summaries improve interpretability of uncertainty regions.

Abstract

Data-driven approaches to predict-then-optimize decision-making problems seek to mitigate the risk of uncertainty region misspecification in safety-critical settings. Current approaches, however, suffer from considering overly conservative uncertainty regions, often resulting in suboptimal decisionmaking. To this end, we propose Conformal-Predict-Then-Optimize (CPO), a framework for leveraging highly informative, nonconvex conformal prediction regions over high-dimensional spaces based on conditional generative models, which have the desired distribution-free coverage guarantees. Despite guaranteeing robustness, such black-box optimization procedures alone inspire little confidence owing to the lack of explanation of why a particular decision was found to be optimal. We, therefore, augment CPO to additionally provide semantically meaningful visual summaries of the uncertainty regions to give qualitative intuition for the optimal decision. We highlight the CPO framework by demonstrating results on a suite of simulation-based inference benchmark tasks and a vehicle routing task based on probabilistic weather prediction.