Global optimization with chaotic particles inspired by swarm intelligence

A new global optimization method utilizing discrete gradient dynamical systems is proposed. The property of the system, including convergence property effective for local search and chaotic property for global search, is subject to a parameter in it, but issues on setting or controlling its value still remain unestablished. We propose an integrated method for updating the parameter's value, in parallel with searching for the global optimizer. Essences to the method are a devised evaluation function for updating values of searching point and parameter, and a practical computational procedure inspired from swarm intelligence. We verify our proposing method with some numerical simulations.