Effect of Local Search on the Performance of Genetic Algorithm

Genetic Algorithms are biologically inspired optimization algorithms that has been used in a number of NP-hard optimization problems successfully like Travelling salesman, Knapsack problem etc. Performance of genetic algorithms largely depends on type of genetic operators Selection, Crossover, Mutation and Replacement used in it. Replacement operator decides which individuals stay in a population and which are replaced by removing or replacing some offspring or parent individuals. Genetic algorithms mainly lead to the premature convergence that makes them incapable of finding global optimal solution. A hybrid algorithm is an extension of genetic algorithm that incorporates the local search techniques within genetic operations so as to prevent the premature convergence. In this paper author is analyzing the effect of hybridization of local search with replacement operators on the performance of genetic algorithm. Implementation is carried out using MATLAB code on test problem Benchmark Dejong's Rastrigin Function (F4). The graphs clearly show that the hybrid algorithm is converging towards optima more quickly than the conventional algorithm. Keywords— Genetic algorithm, hybridization, Genetic and local search algorithm, Dejong function,