Date: Oct 5th, Thursday
Time: 11:45 am
Place: Little 339 (the Atrium)
Speaker: Dzung Phan

Drinks and Pizza will be provided after the talk

Title: Branch and bound approach to globally minimizing
quadratically constrained quadratic programs
　

Abstract

This talk considers the problem of minimizing a quadratic function
subject to infinitely many strictly convex quadratic constraints,
or nonconvex QCQP. In the case where the objective function is
convex, the problem is convex and can be solved efficiently.
Whereas, the nonconvex QCQP is NP-hard, and as a practical matter,
all known algorithms to solve them have a complexity that grows
exponentially with problem dimensions. So it's reasonable to
consider them hard to solve. We propose a branch and bound scheme,
based on a relaxation of the objective function, to globally solve
the nonconvex quadratic programs.