Genetic Algorithms and Neural Networks

摘要: The basic genetic algorithm is introduced including the representation of individuals in populations, data structures for variables, binary strings, assessment individual fitness, selection recombination, crossover and mutation operators. penalty function method handling design constraints introduced. GA illustrated by optimizing a simple structural design. We consider how we might improve on-line adaptation main controls. then review string coding, schema theorem formation building blocks strings. coding continuous-valued variables bit array representations, elitism, methods maintaining diversity population introduce further illustration optimization. application extended into large scale-situations, particularly situations involving number variables. A combinatorial space reduction heuristic based on record parameter intensities described. allocation fitness to partial strings reviewed. Consideration given multi-objective pareto optimality. There follows brief introduction mathematical models GA. used train neural network as an alternative back-propagation. ‘permutations’ problem concept ‘shift’. training analysis. chapter concludes with implicit parallelism suggestions be improved parallel hardware example application.