general constrained optimization;
global convergence;
trust region method;
nonlinear programming;
regularity condition;
D O I:
10.1007/s001860300333
中图分类号:
C93 [管理学];
O22 [运筹学];
学科分类号:
070105 ;
12 ;
1201 ;
1202 ;
120202 ;
摘要:
In this paper, a new trust region method is presented for general constrained optimization problem. In this algorithm, the trial step is obtained by solving two quadratic programming problems with bound constraints. The algorithm is implementable easily. Then we prove that the method is globally convergent without regularity assumptions. Preliminary numerical experiments show the efficiency of the algorithm.