解约束最优化问题的凝聚函数法.docVIP

  An Aggregate Function Method for Constrained Optimization Problems 5 10 15 ZHANG Xinchun* (Research Institute of Electronic Science and Technology, University of Electronic Science and Technology of China, Chengdu 611731) Abstract: In this paper, the aggregate function method (AFM) for constrained optimization problems (COP) is investigated. The AFM is effective for multiple inequality-constrained optimization problems. However, the inherent shortcomings of the AFM, both in convergence and data overflow, seem to have gone unnoticed. To address these issues systematically, some efficient measurements are carried out to deal with such problems, such as adding the stable term and the exponent reduction parameter to the original aggregate function, which are highlighted in the entire paper. Moreover, the corresponding properties of both the AFM and the improved AFM (IAFM) are presented. Key words: Constrained optimization problems; aggregate function; optimization 0 Introduction Consider the typical constrained optimization problem (COP) ? m i n f X( ) ( P)??  i?? 1 , 2?, t ,  (1)  20 25 30 35 40  where f and gi :??n???? are twice continuously differentiable, real and smooth functions, t is the number of the inequality constraints. The core issue of (1) is to solve the multi-constraint inequality optimization problems with the least efforts. Up to now, considerable efforts have been made to deal with these problems. As an ideal candidate, the aggregate function method (AFM) was first formulated and elaborated by Li [1] in 1991. The heart of the AFM is using the maximum” constraint substitute for the original constraint set. Afterwards, the original multi-constrained optimization problem is converted into a single and non-differentiable constrained one. A differentiable function is therefore derived to replace the non-differentiable constraints by using the maximum entropy principle. Namely, (1) is converted to a differentiable optimization problem with the