Learning Optimization Proxies for Sequential Contextual Stochastic Programs: An Order Fulfillment Application
Published in arXiv preprint arXiv:2606.25362, 2026
Tinghan Ye, Shuaicheng Tong, Changkun Guan, Beste Basciftci, Pascal Van Hentenryck. "Learning Optimization Proxies for Sequential Contextual Stochastic Programs: An Order Fulfillment Application." arXiv preprint arXiv:2606.25362, 2026.
Real-time order fulfillment requires solving sequential contextual stochastic programs under tight latency constraints. This paper — the extended version of our NeurIPS 2025 ML×OR workshop paper — develops neural optimization proxies that learn to approximate the solutions of these programs for real-time deployment.
The learned proxies make decisions roughly 2,800× faster than traditional optimization solvers while delivering superior fulfillment cost outcomes and reducing late deliveries by about half compared to established fulfillment policies.