Keywords

oracle learning, perceptrons, neural networks

Abstract

Often the best model to solve a real world problem is relatively complex. The following presents oracle learning, a method using a larger model as an oracle to train a smaller model on unlabeled data in order to obtain (1) a simpler acceptable model and (2) improved results over standard training methods on a similarly sized smaller model. In particular, this paper looks at oracle learning as applied to multi-layer perceptrons trained using standard backpropagation. For optical character recognition, oracle learning results in an 11.40% average decrease in error over direct training while maintaining 98.95% of the initial oracle accuracy.

Original Publication Citation

Menke, J., and Martinez, T. R., "Simplifying OCR Neural Networks with Oracle Learning", Proceedings of the IEEE International Workshop on Soft-Computing Techniques in Instrumentation, Measurement, and Related Applications, pp. 6-13, 23.

Document Type

Peer-Reviewed Article

Publication Date

2003-05-17

Permanent URL

http://hdl.lib.byu.edu/1877/2439

Publisher

IEEE

Language

English

College

Physical and Mathematical Sciences

Department

Computer Science

Share

COinS