Modified Differential Evolution Algorithm with Updated Mutation and Crossover Operator
Surendra Tripathi, Kk Mishra and Shailesh Tiwari
Differential evolution algorithm (DE) is a very popular algorithm which is used to solve numerical optimization problems. The process of DE is implemented with the help of three operators known as mutation, crossover and selection. These operators are designed by mapping the theory of natural selection and evolution. Simplicity of DE motivated many scientists to apply this algorithm for solving real life optimization problems. Although it performed well in solving many optimization problems, yet it was noticed that in some problems, it tends to stick in local optimal solution. Several new variants of DE were developed to solve stagnation problem either by changing the mutation or crossover strategy of DE or by performing parameter tuning. The research of creating fast variant of DE is still on and is dependent on the technique used for implementing the algorithms. Motivating from idea of developing new variant of DE, we did some changes in mutation and crossover operator DE algorithms and created Modified Differential Evolution Algorithm (MDE). We updated mutation operator of MDE to provide more bandwidth for creating diverse solutions and updated crossover operator to make it more effective in exploiting the region around good solutions. The proposed approach not only improves the convergence rate of DE algorithm but is also very useful in maintaining the diversity among solutions. For checking the performance of proposed algorithm, it is compared with other state of art algorithms. The result verifies that proposed approach is good as compared to other latest DE variants.
Keywords: Differential evolution (DE), evolutionary algorithms (EAs), optimization problems, mutation, crossover operator, trail vector