Some Issues On Evolutionary Computation

Evolutionary Computation is the general term for several computational techniques which are based to some degree on the evolution of biological life in the natural world. It's generally considered that EC consists of Genetic Algorithm, Genetic Programming, Evolutionary Strategy. This algorithms has been widely used in machine learning, artificial intelligence, adaptive control, artificial neural network training, Image processing, among other areas.Firstly a new Co-GA model based on binary limited population is presented. Coevolutionary algorithms from the natural world Coevolution mechanism has been widely used in many aspects. Otherwise, It has a weak theory. This model based on Vose's binary limited population model. Convergence analysis and speed-up on this kind of algorithms can be through with this model.The experimental platform for evolutionary algorithms has been discussed in this paper. Particularly a graceful effective framework ,OpenBeagle. The "Population", "Evolver","Internal System" in OpenBeagle has been analysed, and how we construct our own EC algorithms , GA, GP, MOEA, Co-EA based on OpenBeagle. With the help of this good platform ,compare between all kind of EC algorithms could be more easy. The experiments in this paper is based on OpenBeagle.A new genetic algorithm—"Genetic Algorithm Optimal Solution Orientation" is presented. An abstract description of OSOGA has been given. Based on this description a theoretical analysis shows that OSOGA converges under certain conditions. Some Single-objective function optimization problems have been run. The simulation showed that OSOGA can solve the dilemma: premature or non-convergence very well. It can avoid converging to local optimal solution, meanwhile, assemble most computing resources over the sub domain which contains the optimal solution. Theoretical analysis and experiments indicate that OSOGA can find the "optimal" sub domain effectively. Cooperating with local search algorithm, OSOGA can achieve highly precision solution with limited computing resources. In this work some test on Multi-objective function optimization is also been run. The experiments indicates that "OSO Strategy" can speed up the optimization process as a assistant strategy, and improve the distribution of Pareto solution partially.