Machine Learning Inside MIP Solvers - IPAM at UCLA

Explore a comprehensive lecture on the integration of machine learning techniques within Mixed Integer Programming (MIP) solvers. Delve into four key projects that have enhanced the performance of Xpress and SCIP solvers on general MIP benchmarks. Discover how machine learning models are being utilized to make crucial online decisions in various solver subroutines, including presolving, cut generation, cut selection, and primal heuristics. Examine two cutting plane-related topics and two projects focused on improving numerical stability. Gain insights into the challenges of surpassing hand-crafted rules and the growing prominence of machine learning in optimization algorithms. Presented by Timo Berthold from the Technische Universität Berlin at IPAM's Artificial Intelligence and Discrete Optimization Workshop, this 52-minute talk offers a deep dive into the intersection of artificial intelligence and discrete optimization.