Continuous optimization problem declustering algorithm based on double-space cooperation

Abstract

The invention discloses a continuous optimization problem declustering algorithm based on double-space cooperation. Distances between each candidate solution after sequencing according to fitness values and a candidate solution with a higher fitness value is calculated, and a minimal distance is defined as a relative distance of the solution; a distance threshold is determined; according to a sequencing result of the fitness values, each candidate solution is successively processed; a candidate solution whose relative distance is greater than the distance threshold is set as a class center; and class labels are distributed to residual candidate solutions. According to the invention, a class quantity does not have to be specified in advance, the class quantity of clustering is obtained through self-adaption of data features of actual problems, and the adaptability for different distribution is good. The process and the result of the method better accord with features of actual problems, reasonable segmentation of solution space is realized, distribution modes and rules of data are found, and important information is provided for subsequent solving of a continuous optimization problem.

Description

technical field [0001] The invention belongs to the field of intelligent computing and machine learning, and mainly clusters the candidate solutions of continuous optimization problems, and obtains the distribution characteristics of these solutions, so as to assist in improving the performance of intelligent optimization algorithms, and specifically relates to a continuous optimization based on dual-space collaboration Problem Declustering Algorithms. Background technique [0002] At present, intelligent optimization algorithms such as differential evolution, genetic algorithm, particle swarm optimization algorithm, etc. have been widely used in solving continuous optimization problems. Satisfactory results have been achieved for single-mode problems. However, for problems with complex spatial distribution and multi-modes, the existing intelligent optimization algorithms tend to converge prematurely, concentrating on a certain fixed solution area prematurely, and the perfo...

AI Technical Summary

Problems solved by technology

On the one hand, the algorithm needs to calculate the density of each sampling point, here it is necessary to set the parameter radius in advance, and then determine the density according to the distance between points; on the other hand, the remaining two key parameters are determined by the sampling point As shown in the figure, it can be observed that these two points make the artificially determined parameters have a greater impact on the clustering results

Moreover, this method only uses the information of the candidate solution itself, but does not consider the information of the fitness value of the target space.

Images