作者: Philip E. Gill , Walter Murray , Michael A. Saunders , Margaret H. Wright
DOI: 10.1007/BF01589114
关键词: Constrained optimization 、 Linear programming 、 Numerical analysis 、 Set (abstract data type) 、 Simplex algorithm 、 Mathematics 、 Mathematical optimization 、 Sequence 、 Degeneracy (mathematics) 、 Reliability (computer networking)
摘要: A procedure is described for preventing cycling in active-set methods linearly constrained optimization, including the simplex method. The key ideas are a limited acceptance of infeasibilities all variables, and maintenance “working” feasibility tolerance that increases over long sequence iterations. additional work per iteration nominal, “stalling” cannot occur with exact arithmetic. method appears to be reliable, based on computational results first 53 linear programming problems theNetlib set.