Neur2SP- Neural Two-Stage Stochastic Programming - IPAM at UCLA

Explore a cutting-edge approach to solving two-stage stochastic programming problems in this 50-minute lecture by Elias Khalil from the University of Toronto. Delve into the Neur2SP method, which uses neural networks to approximate the expected value function in stochastic programming models. Learn how this innovative technique can efficiently handle mixed-integer linear programs and nonlinear programs in the second stage without relying on problem-specific structure. Discover the advantages of Neur2SP in terms of solution quality and computational speed across various benchmark problems. Gain insights into the method's ability to find high-quality solutions quickly, even as the number of scenarios increases. Compare Neur2SP's performance to traditional solution techniques and understand its potential applications in decision-making under uncertainty.