A Genetic Algorithm for Solving Multimodal Functions Based on Neighborhood Penalty Function

By utilizing the neighborhood penalty function and mutation method, the research puts forward a novel genetic algorithm (GA) by combining global search and local search. Based on the strategy of multiple evolutions, the algorithm constructs a neighborhood with the result of each evolution as the centre, and then sets a penalty function to punish individuals in the neighborhood. The experiment proves that the algorithm converges rapidly, shows favorable global superiority, and is not likely to get trapped in a local optimum. Endowed with these advantages, the algorithm presents preferable global performance and therefore is universally applicable to multimodal functions with multiple solutions.