Prescribe-then-Select: Adaptive Policy Selection for Contextual Stochastic Optimization

2025/09/09 by C Iglesias, Iglesias, Caio de Prospero, Kimberly Villalobos Carballo +3
Computer Science · Decision Sciences · #Advanced Multi-Objective Optimization Algorithms #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Risk and Portfolio Optimization #Stochastic Gradient Optimization Techniques

paper · pdf · doi:10.48550/arxiv.2509.08194

openalex publication_date 2025/09/09 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We address the problem of policy selection in contextual stochastic optimization (CSO), where covariates are available as contextual information and decisions must satisfy hard feasibility constraints. In many CSO settings, multiple candidate policies--arising from different modeling paradigms--exhibit heterogeneous performance across the covariate space, with no single policy uniformly dominating. We propose Prescribe-then-Select (PS), a modular framework that first constructs a library of feasible candidate policies and then learns a meta-policy to select the best policy for the observed covariates. We implement the meta-policy using ensembles of Optimal Policy Trees trained via cross-validation on the training set, making policy choice entirely data-driven. Across two benchmark CSO problems--single-stage newsvendor and two-stage shipment planning--PS consistently outperforms the best single policy in heterogeneous regimes of the covariate space and converges to the dominant policy when such heterogeneity is absent. All the code to reproduce the results can be found at https://anonymous.4open.science/r/Prescribe-then-Select-TMLR.

Citations

• Collective Wisdom: Policy Averaging with an Application to the Newsvendor Problem
• A Survey of Contextual Optimization Methods for Decision Making under Uncertainty
• Multistage Stochastic Optimization via Kernels
• Integrated Conditional Estimation-Optimization
• Optimal Policy Trees
• Residuals-based distributionally robust optimization with covariate information
• Residuals-based distributionally robust optimization with covariate information
• Fast Rates for Contextual Linear Optimization
• Stochastic Optimization Forests
• Dynamic optimization with side information
• Reinforcement Learning with Competitive Ensembles of Information-Constrained Primitives
• Offline Multi-Action Policy Learning: Generalization and Optimization
• Divide-and-Conquer Reinforcement Learning
• Smart "Predict, then Optimize"
• Assessing the Performance of Deep Learning Algorithms for Newsvendor Problem
• Applying Deep Learning to the Newsvendor Problem
• Robust Sample Average Approximation
• From Predictive to Prescriptive Analytics
• Data-Driven Robust Optimization
• Data-driven robust optimization
• On Distributionally Robust Chance-Constrained Linear Programs
• A Stochastic Approximation Method
• Nearest neighbor pattern classification

Related

• A Survey of Contextual Optimization Methods for Decision Making under Uncertainty
• Robust Data-driven Prescriptiveness Optimization
• A Reinforcement Learning Method for Environments with Stochastic Variables: Post-Decision Proximal Policy Optimization with Dual Critic Networks
• Primal-dual algorithm for contextual stochastic combinatorial optimization
• Contextual Stochastic Bilevel Optimization