A multidimensional descent method for global optimization

摘要: This article presents a new multidimensional descent method for solving global optimization problems with box-constraints. is hybrid where local search used and further on the subsets of intersection cones generated by feasible region. The discrete gradient cutting angle search. Two- three-dimensional are Such an approach allows one, as rule, to escape minimizers which not ones. proposed strong properties. We present results numerical experiments using both smooth non-smooth test problems. These demonstrate that algorithm one find or near minimizer.