Numerical tests for Comparision of SeDuMi and SS to solve second order cone programming problems

Abstract

Nonconvex Quadratic Optimization Problem (QOP)은 SDP relaxation과 SOCP relaxation을 통하여 근사해를 찾는다. SOCP relaxation은 SDP relaxation보다 좋은 efficiency를 가진다. SOCP relaxation의 반복 적용함으로 SOCP relaxation의 effectiveness를 향상시킬 수 있다. SOCP relaxation의 반복 적용은 software package를 이용한다. 이 논문에서는 SOCPS의 근사해를 구하기 위한 software package인 SeDuMi (Self-Dual-Minimization)과 SS (SOCP Solver)를 이용하고 몇 가지 수치적 실험을 통해 SeDuMi와 SS를 비교하였다.;Nonconvex Quadratic Optimization Problems (QOPs) are solved approx- imately by SDP (semidefinite programming) relaxation and SOCP (second order cone program) relaxation. SOCP showed efficiency over SDP relaxation, however, its ability to obtain bounds is inferior to SDP. To improve effectiveness of SOCP, iterative application of SOCP is necessary using soft- ware packages. We show numerical tests to compare performance of SeDuMi (Self-Dual-Minimization) and SS (SOCP Solver) for SOCPs.