Comparison.docx

Comparison of Genetic Algorithms and Particle Swarm Optimisation: Ebehart and Yuhushi in 2015 found that the three focused classes of Genetic Algorithms such as Selection, Crossover and And-Mutation can be implied and executed in various ways. On the other hand, Particle Swarm Optimisation lacks such specific operations in that way but the analogies are present. These analogies are depending upon the implementation of Genetic Algorithms’ Operations (Ebehart and Yuhushi, 2015). However, a study conducted by Ebehart, Dobbins and Simpson in 1996 already mentioned the impacts of these operations which choose to get different course for a Run. A Run is the total number of generations GA has got before it’s the operation ends or terminates. The termination occurs because of the excess or maximum numbers of the generations (Ebehart, Dobbins and Simpson, 1996). Genetic Algorithm Crossover impacts in the form of variation significantly during the operation is running. The population members are to be randomized in the start in order to have the prominent impact by the crossover which results in a huge distance in problem space. In 2014, a research conducted by Singh, Kaur and Sinha has shown that the Particle Sworn Optimisation does not include crossover operation however each particle in PSO has been accelerated clearly towards its best position of its entire population which depends upon the class of PSO being used in that operation. These particles are intended to find or range the specific area for a short time that is the geometric mean between two promising areas. Because of the geometric behaviour of PSO, it is more analogous to the recombination for evaluating the strategies (Singh, Kaur and Sinha, 2014). Thus, this analogy is further supported because of the occurrence of recombination on the basis of parameter-by-parameter process. The proper implementation and execution were studied by Shabir and Singla in 2014. According to them, GA is based upon the formation of members to be according to the fitness function predefined by the end of every iteration. After the initiation of GA parameters, three above mentioned main operators of GA (Selection, Crossover and Mutation) are implemented. Then, Fitness evaluation is done on the basis of Benchmark Function and the function is terminated on that criteria. Shabir and Singla also found that PSO on the other hand is an evolutionary algorithm which needs the formation of random numbers and the performance of this algorithm can be affected by the numbers generated and the quality of the operator PSO being used. In this, parameters are to be set up after initiation of the operator. For every particle generated, initial velocity as well as position are to be generated and the fitness value is to be optimized. Local best (pbest) and Global best (gbest) are updated for each particle (Shabir and Singla, 2014). Talking about the comparison, Holland in 1973 found that GA had a nature of discreetness such as it converts the binary functions in to 0s and 1s. as a result, it is not that much difficult to handle the discrete problems. On the other hand, PSO is continuous in nature that requires and modified way to handle the discrete problems. The variables in PSO are able to contain any supposed value based upon the current defined position in the problem space and same is the case with its velocity value. Holland also studied that unlike PSO, GA could not cope with the complexity level in a fluent and efficient way because of the undergone mutation of

the number elements which results in an increased search space significantly. Thus, PSO is a great option when small parameter along with lower number of iterations are required (Holland, 1973). PSO always work for the global optima and tries to search the global points unlike GA which mainly converges towards the local medium of optima finding the arbitrary points. Mendes in his Phd dissertation in 2004 researched that the Particle Swarm Optimization and Genetic Algorithm both change on the basis of the set of points with in an iteration having a visible improvement comparing previous values. Although, both GA and PSO are the basic parts of the optimization, they are not perfect thus have limits in their utility to only few problems.

References: Russell C. Eberhart and YuhuiShi, 2015, Comparison between Genetic Algorithms and Particle Swarm Optimization. Department of Electrical Engineering Indiana University Purdue University Indianapolis 723 W. Michigan St., SL160 Indianapolis, IN 46202 Eberhart, R. C., Dobbins, R. W., and Simpson, P. K. 1996, Computational Intelligence PC Tools, Boston: Academic Press. S. Singh, J. Kaur, and R. S. Sinha, 2014, A comprehensive survey on various evolutionary algorithms on gpu. Shahid Shabir and Dr. Ruchi Singla, 2016, A Comparative Study of Genetic Algorithm and the Particle Swarm Optimization. International Journal of Electrical Engineering. ISSN 0974-2158 Volume 9, Number 2 (2016), pp. 215-223 J. H. Holland, ―Genetic algorithms and the optimal allocation of trials, ‖ SIAM Journal on Computing, vol. 2, no. 2, pp. 88–105, 1973. R. Mendes, ―Population topologies and their influence in particle swarm performance, Ph.D. dissertation, 2004