Hybrid genetic algorithm for fast electromagnetic synthesis

Paper describes a hybrid genetic algorithm (HGA) developed as a combination of a binary genetic and steepest descent gradient (SDG) algorithms. Here genetic algorithm (GA) performs global exploration of the whole design space aiming at identification of a given number of promising solutions, whereas SDG algorithm is used to refine these nearly optimal solutions via local down-hill optimization. Such a two-step strategy enables one to significantly reduce stagnation often observed for GAs at the later stage of optimization. In particular it is demonstrated that the proposed HGA has twice faster convergence rate comparing to standard GAs and much better stability in hitting global minima. The latter is strongly facilitated by implementation of advanced encoding schemes.