A challenge to huge scale semidefinite programs-exploiting sparsity, parallel computation and polynomial optimization problems

半定规划是在科学和工程领域有着广泛应用的优化问题之一，我们研究了半定规划的原-对偶内点方法，并在此基础上开发了一个软件包SDPA。与现有的软件包相比，SDPA在更短的时间内解决了更大规模的问题。作为SDPA的应用，我们为多项式优化问题提供了SparsePOP，为大规模传感器网络定位问题提供了SFSDP。

We studied primal-dual interior-point methods for solving a semidefinite program which is one of the most important optimization problems having lots of applications in various fields of science and engineering, and developed a software package SDPA based on them. SDPA solves larger scale problems in shorter time than the existing software packages. As applications of SDPA, we provided SparsePOP for polynomial optimization problems and SFSDP for large scale sensor network localization problems.