Particle Swarm Optimization with inertia non-monotonic control

Abstract

This work presents a mechanism to reduce the chances of the optimization process of nonlinear functions stagnating in local minima, using the meta-heuristic “Particle Swarm Optimization”. This mechanism is a nonmonotonic way to control the particle inertia, which is one of the factors responsible for movement during the optimization process. The experimental results were compared to the original PSO model aiming to show the potential to find a better solution related to the benchmark functions for complex problems. Finally, a comparison of both models was made in the adjustment of synaptic weights of a Multi-Layer Perceptron neural network obtaining interesting results.

Keywords: swarm intelligence, Particle Swarm Optimization, artificial life, artificial neural network MLP.