Presented and published in the proceedings of the
International Joint Conference on Intelligent Systems (JCIS'00)
Atlantic City, NJ, Feb.2000, pp.887-891.

 

Combining Constructive Neural Networks for Ensemble Classification

 

R.S. Renner
Department of Computer Science
California State University, Chico
Chico, CA 95929-0410 USA

R.C. Lacher
Department of Computer Science
Florida State University
Tallahassee, FL USA

 

ABSTRACT
Ensemble networks have been receiving considerable attention within the last few years. Most existing models are created with linear networks. Ensembles of linear networks have demonstrated improved performance over individual networks, but linear models have limited capacity. Ensembles of more complex well-trained networks offer an alternative. Unfortunately, the computational expense involved in training large numbers of well-trained networks is often prohibitive. An ensemble model consisting of Cascade Correlation neural networks is presented. Cascor ensembles exhibit improved generalization over ensembles of linear students without prohibitive time-complexity. An ensemble combination method based on validation error is introduced as an alternative to voting and simple averaging. The Neural Network Ensemble Simulator (NNES) is introduced.

Keywords: neural network ensembles, Neural Network Ensemble Simulator, NNES, ensemble combination.


* Present Address: Dept. of Computer Science
California State University - Chico
Chico, CA 95929-0410
E-mail: renner@ecst.csuchico.edu