Summary

International Symposium on Nonlinear Theory and Its Applications

2022

Session Number:C3L-C

Session:

Number:C3L-C-01

Learning a Simple Multilayer Perceptron with PSO

Riku Takato ,   Kenya Jin'no,  

pp.470-473

Publication Date:12/12/2022

Online ISSN:2188-5079

DOI:10.34385/proc.71.C3L-C-01

PDF download (2.3MB)

Summary:
In this article, we attempt to learn the parameters of a multi-layer perceptron (MLP) using the particle swarm optimization (PSO) method which is one of the approximate solution methods for optimization problems without using the derivative information of the objective function. We use gradient method and PSO to learn to classify a linearly inseparable data set with an MLP in the middle layer with a small number of neurons. As a result, we experimentally confirm that PSO outperforms gradient-based learning.