Improving the performance of cascade correlation neural networks on multimodal functions

Riley, Mike and Jenkins, Karl W. and Thompson, Chris P. (2010) Improving the performance of cascade correlation neural networks on multimodal functions. In: World Congress on Engineering 2010, 30 June - 2 July 2010, London, UK.

Full content URL: http://www.iaeng.org/publication/WCE2010/WCE2010_p...

Documents
WCE2010_pp1980-1986.pdf
[img]
[Download]
[img]
Preview
PDF
WCE2010_pp1980-1986.pdf - Whole Document

1MB
Item Type:Conference or Workshop contribution (Paper)
Item Status:Live Archive

Abstract

Intrinsic qualities of the cascade correlation algorithm make it a popular choice for many researchers wishing to utilize neural networks. Problems arise when the outputs required are highly multimodal over the input domain. The mean squared error of the approximation increases significantly as the number of modes increases. By applying ensembling and early stopping, we show that this error can be reduced by a factor of three. We also present a new technique based on subdivision that we call patchworking. When used in combination with early stopping and ensembling the mean
improvement in error is over 10 in some cases.

Keywords:Neural Networks, Response Surface, Early Stopping, ensemble machine learning
Subjects:G Mathematical and Computer Sciences > G760 Machine Learning
Divisions:College of Science > School of Engineering
Related URLs:
ID Code:13769
Deposited On:10 Apr 2014 12:08

Repository Staff Only: item control page