Convex Optimization-Based Procedures for Non-Convex Quadratic Problems
A broad family of practical design problems can be represented using quadratically constrained quadratic programs (QCQPs), where both the objective function and the constraints are quadratic functions of the optimization variables.