Population Initialization Methods for the Swallow Swarm Algorithm in Solving the Problem of Fuzzy Classifier Parameter Optimization
Артем Слезкин, I. A. Hodashinsky · 2021
This paper considers and compares population initialization methods for the swallow swarm algorithm in solving the problem of the fuzzy classifier parameter optimization.Population initialization is important in swarm and evolutionary optimization algorithms, in which the lack of diversity in the population can lead to early convergence and to hitting the local optimum.Methods based on quasi-random sequences, chaotic maps and random value distributions were considered.The hybrid initialization method based on the normal and uniform distribution revealed the lowest classification error.The fastest convergence was shown by the method based on the beta distribution.