An Irregularity Measure of Cost Landscape Using Cost Diagonal Patterns Generated in Search with eDEAS

This paper proposes a new method of measuring irregularity of cost landscape by using cost diagonal patterns generated during search with eDEAS that is acknowledged as one of the fastest global optimization methods. This is a preliminary study for developing intelligent dynamic encoding algorithm for searches (iDEAS) which alternates DEAS patterns among eDEAS, uDEAS, and mDEAS according to the measured landscape irregularity. Suitability of the proposed measure is validated through four representative test functions with good agreement. In addition, the proposed measure provides information on dominance of search variable on the cost function, which is helpful to determine the most important variable first in optimal design area.