A parallel implementation for solving large-scale Semidefinite Programs having sparse Schur complement matrix

减少求解半定程序的计算时间对于传感器网络定位问题和多项式优化等许多应用是必不可少的。我们重点研究了占用大部分计算时间的Schur补矩阵的稀疏性。我们估计了矩阵中每个元素的计算成本，并提出了有效的并行方案。数值结果表明，该并行方案有效地减少了计算时间。

The reduction in computation time for solving Semidefinite Programs is essential to many applications like sensor network localization problems and polynomial optimizations. We focus on the sparsity of the Schur complement matrix, which occupies most computation time. We estimate the computation cost of each element of the matrix and propose efficient parallel schemes. From numerical results, we verified that the parallel schemes successfully reduce the computation time.