Research On The Improvement Swarm Intelligence Based On Multidimensional Measurable Space

The rapid development of socio-economic science has made human life more colorful.In life,the problems encountered with various fields also gradually become more complex and diverse.Traditional methods of dealing with problems have been unable to cope with the general needs of modern society,especially in the field of engineering science and technology.For multi-dimensional complex problems that are difficult to solve,people are increasingly demanding the performance of optimization techniques.The traditional optimization algorithms such as linear programming,Newton method,etc.are not satisfactory when solving the high-dimensional,multi-peak and other problems encountered in various fields,however,the swarm intelligent algorithm has good global optimization and gets attention gradually.This paper mainly introduces three algorithms of particle swarm,fruit fly,and flower pollination.Based on this,the main improvements and innovations of this paper are as follows:1)Design a coordinated dynamic learning factor for the particle swarm algorithm so that the best point of self-recording is considered in the initial iteration process,and converges to the optimal point of the population quickly in the later stage of the algorithm.At the same time,in order to overcome the premature phenomenon,when the variance of population is judged by the variance of fitness,the chaos map is used to update the individual and select it again and perform the optimization operation in a new way.2)For the limitations of fruit fly algorithm that only converge to the optimal individuals,the escape mechanism of wavelet transformation is added to ensure the correctness of the selection of iterative direction.When the population diversity is low,the population is subjected to inverse wavelet transformation,which directs the population to escape from the local restricted area and converges to the global optimal solution.3)Adding elite individuals to increase population diversity and using a linearly decreasing traction factor to induce elite individuals to collaborate on optimization from the early stages of the algorithm,it expands its global search capabilities.When the degree of individual aggregation in the later stage of the algorithm becomes larger and the individual diversity of fruit fly becomes lower,the searching strategy of search space compression is introduced to dynamically change the spatial domain of the target problem into an adaptive step length method to help the algorithm jump out of the local position.In-depth optimization.4)Analyzing the constant step size of fruit fly algorithm will affect the optimization accuracy of the algorithm.The iterative step value of the algorithm is used as the guide factor to design an adaptive search method to coordinate the global search ability and the local search ability.In the late stage of algorithm searching,in order to avoid the premature loss of population diversity and resulting in the local optimal solution of the final problem,the cloud model is used as the basis to design the cloud escape mechanism to assist the algorithm to jump out of the local limit for deep search.5)The introduction of the cosine function as the control factor in the flower pollination algorithm randomly adjusts the evolution ratio between the current position and the optimal position,and improves the possibility that the algorithm is too early to gather in the optimal individual and leads to poor evolution.In order to restrain the phenomenon of optimal decomposing and reduce the diversity of the population at the later stage of evolution,a non-uniform mutation strategy was introduced to change the current position to expand the new evolution direction.Simulation experiments show that the improved algorithm has better performance in high-dimensional functions and neural network optimization.