Abstract
In this paper, the performance of two commonly used associative memories, generalised inverse (GI) and correlation memory matrix (CMM), is studied analytically and compared. The effects of crosstalk and a new type of noise (memory noise appearing due to noisy training patterns) are also considered. Theoretical expressions are determined for the S N R (signal to noise ratio) gain of the GI and C M M memories in the auto-associative and hetero-associative modes of operation. It is found that the GI method performance degrades significantly in the presence of memory noise while the C M M method is relatively unaffected by it. In overall performance the GI method is superior to that of the C M M method except at high values of input stimulus noise or when the number of associations stored is close to the dimension of the key vectors. The theoretical expressions are plotted and compared with the results obtained from Monte Carlo simulations and the two are found to be in excellent agreement.
| Original language | English (US) |
|---|---|
| Title of host publication | Proceedings - 1992 International Joint Conference on Neural Networks, IJCNN 1992 |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 184-189 |
| Number of pages | 6 |
| ISBN (Electronic) | 0780305590 |
| DOIs | |
| State | Published - 1992 |
| Event | 1992 International Joint Conference on Neural Networks, IJCNN 1992 - Baltimore, United States Duration: Jun 7 1992 → Jun 11 1992 |
Publication series
| Name | Proceedings of the International Joint Conference on Neural Networks |
|---|---|
| Volume | 2 |
Conference
| Conference | 1992 International Joint Conference on Neural Networks, IJCNN 1992 |
|---|---|
| Country/Territory | United States |
| City | Baltimore |
| Period | 6/7/92 → 6/11/92 |
Bibliographical note
Publisher Copyright:© 1992 IEEE
Keywords
- Associative Memory
- Correlation Memory Matrix
- Generalized Inverse
- Noise Performance