Incremental growth in modular neural networks.

MacLeod, Christopher; Maxwell, Grant; Muthuraman, Sethuraman

Abstract

This paper outlines an algorithm for incrementally growing Artificial Neural Networks. The algorithm allows the network to expand by adding new sub-networks or modules to an existing structure; the modules are trained using an Evolutionary Algorithm. Only the latest module added to the network is trained, the previous structure remains fixed. The algorithm allows information from different data domains to be integrated into the network and because the search space in each iteration is small, large and complex networks with a modular structure can emerge naturally. The paper describes an application of the algorithm to a legged robot and discusses its biological inspiration.

Citation

MACLEOD, C., MAXWELL, G. and MUTHURAMAN, S. 2009. Incremental growth in modular neural networks. Engineering applications of artificial intelligence [online], 22(4-5), pages 660-666. Available from: https://doi.org/10.1016/j.engappai.2008.11.002

• Journal Article Type: Article
• Acceptance Date: Nov 7, 2008
• Online Publication Date: Dec 21, 2008
• Publication Date: Jun 30, 2009
• Deposit Date: Jul 24, 2009
• Publicly Available Date: Jul 24, 2009
• Journal: Engineering applications of artificial intelligence
• Print ISSN: 0952-1976
• Electronic ISSN: 1873-6769
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 22
• Issue: 4-5
• Pages: 660-666
• DOI: https://doi.org/10.1016/j.engappai.2008.11.002
• Keywords: Genetic algorithm; Modular artificial neural network; Robotics; Incremental growth; Incremental evolution; Evolutionary algorithm
• Public URL: http://hdl.handle.net/10059/382
• Contract Date: Jul 24, 2009