Speaker: Burak Kocuk ( Associate Professor, Industrial Engineering Program at Sabancı University)
Title: Learning to Relax Nonconvex Quadratically Constrained Quadratic Programs
Date: 1st of November , 2024 (Friday)
Time: 13:30-14:30
Place: EA409
Abstract: Nonconvex quadratically constrained quadratic programs are NP-Hard to solve in general. Literature primarily employs either semidefinite or linear programming to relax them, which are usually effective for distinct problem types, and a holistic understanding of which relaxation should be preferred over the other for a given instance is lacking. In this research, we present a learning-based approach to predict whether a semidefinite or linear programming relaxation would produce a stronger bound for a given instance by examining spectral properties and sparsity patterns of the data matrices.
Bio: Burak Kocuk is an associate professor in the Industrial Engineering Program at Sabancı University. He obtained his BS degrees in Industrial Engineering and Mathematics, and MS degree in Industrial Engineering from Boğaziçi University. He obtained his PhD degree of Operations Research from the School of Industrial and Systems Engineering at Georgia Institute of Technology. Before joining Sabancı University, he was a postdoctoral fellow in the Tepper School of Business at Carnegie Mellon University. His current research focuses on mixed-integer nonlinear programming and stochastic optimization problems, from both theoretical and methodological aspects.