Speciation-based genetic algorithm in analog circuit design

This paper presents a speciation procedure that improves the local search capability of the genetic algorithm in analog circuit design. There is no need for additional circuit simulation in order to apply this procedure. The procedure is tested in Gaussian, sigmoid, cube, and square circuit design problems. Two sets of 125 simulations with the same seed values are performed for each problem using both the proposed procedure and the canonical genetic algorithm. The simulation results show that the method is statistically better than the canonical genetic algorithm, which suffers from bad locality. The effects of the population size and speciation threshold coefficient on the performance of the speciation algorithm are investigated. Confidence intervals of the simulation results are calculated. The results show that the speciation procedure improves the quality of solutions with at least 99% confidence, and the effectiveness of the method, which is statistically determined, increases in small populations