Proceedings of Navy SBIR Neural Network Conference
Abstract
This volume contains sixteen contributions by invited representatives to a Neural Network Applications Conference sponsored by the Navy's Small Business Innovation Research (SBlR) Program. The Office of Naval Research (ONR) was the first major governmental organization to have a program dedicated to the basic science of understanding neural networks. The program began six short years ago and quite soon it became evident that Navy-relevant applications were possible. The past six years have shown that neural networks have progressed from being great ideas in their creators' heads to engineered applications that show results. In the beginning there was no clear winner in the Great Best Network Model Race, but the papers in this volume reveal that "back propagation" (in its various forms) has pulled far ahead of the pack. I would suggest the following reasons: Back propagation training is mathematically organized to produce a minimization of the mean squared aggregate error across training patterns. Applications oriented engineers and small business entrepreneurs are comfortable with LMS error techniques, and back propagation is able to minimize the mean squared error of the training data with relatively simple mathematical proofs.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1992
- Accession Number
- ADA401420