1993 Convergence behavior of interior-point algorithms文献.pdfVIP

Mathematical Programming 60 ( 1993) 215-228 215 North-Holland Convergence behavior of interior-point algorithms Osman Gfiler and Yinyu Ye* Department of Management Sciences, The University of Iowa, Iowa City, IA, USA Received 12 February 1991 Revised manuscript received January 1992 We show that most interior-point algorithms for linear programming generate a solution sequence in which every limit point satisfies the strict complementarity condition. These algorithms include all path-following algorithms and some potential reduction algorithms. The result also holds for the monotone complementarity problem if a strict complementarity solution exists. In general, the limit point is a solution that maximizes the number of its nonzero components among all solutions. AMS Subject Classifications: Primary 90C05, 90C33. Key words: Strict complementarity, maximal complementarity, interior point algorithms, linear program- ming, monotone complementarity problem. 1. Introduction Since K a r m a r k a r [13] p u b l i s h e d his p o l y n o m i a l - t i m e algorithm for linear program- ming, m a n y interior point algorithms have been p r o p o s e d for special classes of convex programs. These algorithms attempt to solve a given p r o g r a m by iteratively generating interior points which converge to a solution of the p r o b l e m in p o l y n o m i a l time. Polynomial-time interior point algorithms have been found for linear program- ming, convex quadratic p r o g r a m m i n g , and also some classes of s m o o t h c o n v e x programs. The more recent of these are p r i m a l - d u a l algorithms which take advantage of duality