Improved Chaotic Differential Evolution Algorithm And Its Application

Chaotic differential evolution algorithm integrates the fine search of chaos optimizationwith fast local search ability and the differential evolution algorithm with fast global searchability. It overcomes the original differential evolution too fast convergence and terminationof the iterative defects. At the same time, the convergence of the algorithm to the optimalsolution is faster and finds the optimal value of accuracy is retained; Chaotic differentialevolutionary algorithm which improved, is mainly to control the parameters value selecting oforiginal algorithm intelligently, and joins the precision control factor. It not only retains thesuperiority of the original chaotic differential algorithm, but also reduces the variableparameter operation intelligently, making the actual use more convenient. The numericalexperiment results show that, by using this improved algorithm for solving several classes o fstandard test functions, the optimization effect is better and the time is also very modest.The chaotic differential evolution algorithm in the application of the improved complexcombinatorial optimization problem, the optimization effect is very good. In view of theactual processing problems for constrained multi-objective problems, the judgments ofconstraint condition are added to each algorithm iteration solution individually. It must repeatagain in the feasible region if the individual doesn’t satisfy these constraints, until it meets allconstraints. In order to enhance the processing of multi constraint, the algorithm can also do ajob with skill and ease in handling complicated combinatorial optimization problems.Multicolored programming optimization problems, such as the traveling salesman problemand the0-1knapsack, can be worked out quickly by the reasonable model generated by thisalgorithm.Combined chaotic differential evolution algorithm with BP neural network in theapplication of IRIS data set classification, the optimization effect is very significant. Theoriginal algorithm with BP neural network and it’s modified algorithm are converging slowly,realizing difficultly and falling into local optimum easily. To avoid these disadvantages, a newhybrid intelligent algorithm is formed by using the improved chaotic differential evolution forupdating gradient threshold value instead of the original BP network with the descent strategy.The experimental results show that, in the IRIS data classification with this algorithm, itsperformance is stronger than the standard BP neural network, the optimization of the PSOneural network and GA neural network optimization.