Abstract
Many artificial neural networks in practical use can be demonstrated not to be fault tolerant; this can result in disasters when localized errors occur in critical parts of these networks. In this paper, we develop methods for measuring the sensitivity of links and nodes of a feedforward neural network and implement a technique to ensure the development of neural networks that satisfy well-defined robustness criteria. Experimental observations indicate that performance degradation in our robust feedforward network is significantly less than a randomly trained feedforward network of the same size by an order of magnitude.
Original language | English (US) |
---|---|
Title of host publication | IEEE International Conference on Neural Networks - Conference Proceedings |
Publisher | Institute of Electrical and Electronics Engineers Inc. |
Pages | 783-788 |
Number of pages | 6 |
Volume | 1993-January |
ISBN (Print) | 0780309995 |
DOIs | |
State | Published - 1993 |
Event | IEEE International Conference on Neural Networks, ICNN 1993 - San Francisco, United States Duration: Mar 28 1993 → Apr 1 1993 |
Other
Other | IEEE International Conference on Neural Networks, ICNN 1993 |
---|---|
Country/Territory | United States |
City | San Francisco |
Period | 3/28/93 → 4/1/93 |
ASJC Scopus subject areas
- Control and Systems Engineering
- Software
- Artificial Intelligence