Characterization of the objective function landscape using a modified Dijkstra algorithm

A decisive disadvantage of high dimensional optimization problems is the large number parameters which make identification process difficult. Many this are reflected in mathematical properties objective function used for parameter identification. Ideally, has a simple, paraboloid-like form with single minimum whose location corresponds to desired set. Instead, unfavorable “landscape forms” occur, example “bumpy” areas several local minima. In literature, dominant methods attempt simplify high-dimensional functions, or research new, mostly stochastic methods. This article examines way obtain information about landscape underlying an problem. path search algorithm based on Dijkstra presented. The purpose find approximately deepest between two minima, allows characteristics function. It tested Himmelblau and Rastrigin examine minima respect their topographic characteristics.