A Penalized Quadratic Convex Reformulation Method for Random Quadratic Unconstrained Binary Optimization

The Quadratic Convex Reformulation (QCR) method is used to solve quadratic unconstrained binary optimization problems. In this method, the semidefinite relaxation is used to reformulate it to a convex binary quadratic program which is solved using mixed integer quadratic programming solvers. We extend this method to random quadratic unconstrained binary optimization problems. We develop a Penalized QCR method where the objective function in the semidefinite program is penalized with a separable term to account for the randomness in the objective.

Article

Download

View A Penalized Quadratic Convex Reformulation Method for Random Quadratic Unconstrained Binary Optimization