In resent years, semidefinite program(SDP)has been intensively studied both in theoretical and practical aspects of various fields including interior-point methods, combinatorial optimization and the control and systems theory. The SDPA(SemiDefinite Programming Algorithm)is an optimization software, implemented by C++ language, of a Mehrotra-type primal-dual predictor-corrector interior-point method for solving the standard form semidefinite program. In this paper, we also discuss parallel execution of the SDPA on the Ninf, a global network-wide computing infrastructure which has been developed for high-performance numerical computation services. We report some numerical results on a parallel implementation of the successive convex relaxation method proposed by Kojima and Tuncelapplying the SDPA on the Ninf.
|Translated title of the contribution||The SDPA (SemiDefinite Programming Algorithm) on the Ninf (A Network based Information Library for the Global Computing)|
|Number of pages||6|
|Publication status||Published - May 25 2001|