A new global optimization algorithm using stochastic model and its application to electronic circuit design

A novel algorithm based on an internal model of an objective function for seeking a global optimum is presented. By minimizing this internal model function, constructed by a stationary Gaussian process, the minima of the objective function can be obtained after an extremely small number of objective function evaluations. The usefulness of this algorithm is demonstrated by several numerical examples and its application to circuit design. The computational overhead for the optimization procedure is low as the model function is based on a stationary Gaussian process.<
>