A distributed algorithm for high-dimension convex quadratically constrained quadratic programs

We propose a Jacobi-style distributed algorithm to solve convex, quadratically constrained quadratic programs (QCQPs), which arise from broad range of applications. While small medium-sized convex QCQPs can be solved efficiently by interior-point algorithms, large-scale problems pose significant challenges traditional algorithms that are mainly designed implemented on single computing unit. The exploding volume data (and hence, the problem size), however, may overwhelm any such units. In this paper, we for general, non-separable, QCQPs, using novel idea predictor-corrector primal-dual update with an adaptive step size. enables storage as well parallel computing. establish conditions proposed converge global optimum, and implement our computer cluster multiple nodes Message Passing Interface (MPI). numerical experiments conducted sets various scales different applications, results show exhibits favorable scalability solving problems.