Particle Swarm Optimizationを用いた階層型ニューラルネットワークの一構成

Design of a Multi-Layered Neural Network Using Particle Swarm Optimization

加藤 暢恵

広島大学大学院教育学研究科紀要. 第二部, 文化教育開発関連領域

Bulletin of the Graduate School of Education, Hiroshima University. Part. II, Arts and science education

Particle Swarm Optimization (PSO), whose concept has been established as a simulation of a simplified social milieu, is known as one of the most useful optimization methods for solving non-convex continuous optimization problems. This paper describes a new learning algorithm to simultaneously adjust connection weights included in neural networks and some user-specified parameters included in units. According to the proposed algorithm, it is possible to improve the learning properties of the neural networks, e.g., the learning cost and/or adaptability. The behavior of the proposed algorithm is examined on a numerical simulation example.

neural network

evolutionary computation

learning

particle swarm optimization

meta heuristics

ニューラルネットワーク

進化計算

学習

PSO

メタヒューリスティク

教育 [ 370 ]

日本語

広島大学大学院教育学研究科

[ISSN] 1346-5554

[NCID] AA11618725