Extensions of  the Semidefinite Coordinate Direction Algorithm for Detecting Necessary Constraints to Unbounded Regions

 

Shafiu Jibrin
Department of Mathematics & Statistics
Northern Arizona University
Flagstaff, AZ 86011

 

Susan Perrone

Department of Mathematical Sciences

Clemson University,

ClemsonSC 29634

 

 

Abstract: In semidefinite programming (SDP) it is highly beneficial to eliminate redundant constraints or identify necessary constraints before solving the problem. Eliminating redundant constraints in SDP is an extremely hard problem. However, there some practical methods for detecting necessary constraints in SDP, including the Semidefinite Stand-and-Hit, Semidefinite Hypersphere Direction, and Semidefinite Coordinate Direction methods. These methods work only with bounded feasible regions. In this talk we give four extensions of the Semidefinite Coordinate Direction method that work in both bounded and unbounded feasible regions.