Multimodal Fish School Search Algorithms Based on Local Information for School Splitting

This work delves into variations of FSS that uses local information (i.e. fish weights) for splitting the school and presents comparative analyses of the new method, tried here in three ways. Hence, this is an attempt to create a more economical alternative for the best performing multimodal version of the algorithm FSS, the dFSS. The work capitalizes on some modifications in the Collective Instinctive Movement and the creation of a simpler function for the links (between fish) definition and put forward in a previous work. The objective was to reduce the number of false-positives without losing the diversity of the solution set returned by the algorithm. The experiments showed that the new algorithm achieved better results considering the following objectives: reduce the number of false-positives, but without losing the ability of finding a diversified set of correct solutions, considering that we did not want to add extra parameters, as well as to perform distance calculations or using any other sort of spatial information. Moreover, the comparative results show that the complexity of the "Guide-Guided" Relationship Links Definition Operator was significantly reduced from O(n
²
) to O(n).