Intelligent Genetic Algorithms in Evolutionary Computation Part ii Application to Combinatorial Multimodal Optimization Problems
Liwen He, Neil Mort · White Rose Research Online (University of Leeds, The University of Sheffield, University of York) · 1998
Combinatorial Multimodal Optimization Problems (CMOP) arising in the scheduling of manufacturing systems involve the determination of multiple integer solution vectors that optimize a given objective function with regard to some definite constraints. The genetic algorithm is an effective computational paradigm to search such a large optimization space for the best solutions. But simple genetic algorithms are notorious for their "premature convergence" to a unimodal (global or local) optimum because of genetic drift. Following the previous discussion in Part 1, a new Intelligent Genetic Algorithm is developed to include spatial structured population, relative fitness vector, absolute fitness value with dynamic fitness sharing function, super conservative selection strategy, intelligent recombination through speciation, optimal outbreeding and neural mutation. Experimental results and simple fitness landscape analysis illustrate that intelligent genetic algorithms can effectively solve a typical combinatorial multimodal optimization problem, which is a challenging problem for GA applications. These advanced intelligent genetic algorithms present a prospective arena for multi-objective optimization [Fonesca and Fleming., 1995a,b; Shaw and Fleming, 1996]and multimodal optimization problems in evolutionary computation.