© The Institution of Electrical Engineers
A RAM-based auto-associative neural network is described which has several desirable properties, including high storage capacity and absence of minima during recall. The system is implemented as a set of generalising RAM (GRAM) type nodes. The generalisation procedure is described, and comparisons with other types of autoassociator are drawn.
References
-
-
1)
-
R.P. Lippmann
.
An introduction to computing with neural nets.
IEEE ASSP Magazine
,
4 -
22
-
2)
-
Kan, W., Aleksander, I.: `A probabilistic logic neuron network for associative learning', Proc. IEEE Int. Conf. on neural networks, 1988, San Diego, p. 541–548.
-
3)
-
Aleksander, I.: `Ideal neurons for neural computers', Proc. of Int. Conf. on Parallel Processing in Neural Systems and Computers, 1990, North Holland Elsevier Science, p. 225–232.
-
4)
-
L. Tarassenko ,
J.N. Tombs ,
J.H. Reynolds
.
Neural network architectures for content addressable memory.
IEEE Proc. F
,
1 ,
33 -
39
http://iet.metastore.ingenta.com/content/journals/10.1049/el_19910501
Related content
content/journals/10.1049/el_19910501
pub_keyword,iet_inspecKeyword,pub_concept
6
6