Random and Chaotic Sequences, and the Effect of their Distributions on PSO Performance
Hendrik Richter, Paul Moritz Nörenberg · 2024
Empirical results show that performance of particle swarm optimization (PSO) may be different if either chaotic or random sequences are used for driving the algorithm's search dynamics. We analyze the phenomenon by evaluating the performance based on a benchmark of test functions and comparing random and chaotic sequences according to equality or difference in underlying distribution or density. Our results show that the underlying distribution is the main influential factor in performance. Thus, it seems not plausible to assume general and systematic performance differences between chaos and random.