Improved Lévy Flight Distribution For High-dimensional Numerical Optimization Problems

The Lévy Flight Distribution Algorithm(LFD)is a new kind of population based meta-heuristic algorithm.LFD is inspired by random walk behaviour of Lévy flight,and has attracted the attention of researchers because of its simplicity and effectiveness.The simulation results show that this algorithm has better overall performance than the existing partial evolutionary algorithms,but like other classical meta-heuristic algorithms,it also has many shortcomings,such as,the lack of population diversity is easy to fall into the local optimal exploration and mining ability imbalance.In addition,with the development of science and technology,the dimension of actual optimization problems is getting higher and higher,and the existing meta-heuristic algorithm is difficult to solve such problems.Therefore,it is necessary to further improve the ability of algorithms to solve high-dimensional problems.In view of these problems,the following works have been done in this dissertation.(1)A Modified Levy Flight Distribution(MLFD)is proposed in this dissertation.In MLFD,in order to improve the diversity of the initial population and better solve the high-dimensional numerical optimization problem,the chaotic mechanism is introduced into the initialization stage.Meanwhile,the mutualism symbiosis stage of the symbiosis organisms search algorithm is introduced to balance the global exploration and local mining ability of the algorithm.In addition,in order to help the algorithm escape the local minimum region in time,an adaptive neighbourhood search operator is introduced to enhance the diversity of the population at the end of iteration.Finally,the feasibility and effectiveness of the proposed MLFD algorithm are verified by 19 well-known largescale unconstrained problems.The experimental results and statistical analysis show that the proposed MLFD algorithm has significant advantages over the other nine classical evolutionary algorithms in high-dimensional numerical optimization problems.(2)In order to overcome shortcomings of LFD algorithm and further improve its ability to solve numerical optimization problems,an multiple sub-population LFD named MSLFD is proposed.Firstly,the marine predator search strategy is introduced to improve the ability of LFD algorithm to solve high dimensional optimization problems.At the same time,the multiple sub-population strategy is adopted to further improve the convergence precision and speed of the algorithm.In addition,in order to avoid the algorithm trapped in local minima area,a improved the neighbourhood search strategy applied to the late iteration of the algorithm.Finally,through the use of 19 famous large-scale unconstrained problems are utilized to confirm the performance of the MSLFD algorithm compared to the state-of-the-art 9 algorithms.Experimental results and statistical analysis demonstrate that the MSLFD algorithm has better performance than other nine classical evolutionary algorithms.