Dynamic tunneling algorithm for global optimization

Abstract

A dynamic tunneling algorithm for global optimization, which is motivated by the concept of dynamic flow and tunneling phenomena, is proposed. The algorithm requires the memory of the most recently found minimum and the focal smoothness in the energy landscape. It is based on two systems: a dynamic optimization system by which a local minimum is found, and a dynamic tunneling system by which a new initial condition in a 'lower valley' is determined. The computation is based on every equilibrium dynamic system always going to the equilibrium to which its initial condition belongs. Numerical results for several typical examples are presented, which clearly illustrate the efficiency of the algorithm.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>