Abstract
The performance of Hebbian-type associative memories (HAM's) in the presence of faulty interconnections is examined and equations for predicting network reliability are developed. The interconnection faults considered are the equivalents of open-circuit and short-circuit synaptic interconnections (analogously, opaque or clear weight mask elements in optical implementations). Our results show that a network with open-circuit interconnection faults has a higher probability of direct convergence, Pdc, than a network with short-circuit interconnection faults when the fraction of failed interconnections, p, is small and the short-circuit signal, G, is large. Our results are also extended to the case where network attraction radius is considered. During the analysis, a HAM is assumed to have a high probability of correct recall; i.e., memory storage is within its capacity. The individual neuron updating in an asynchronous network is assumed to occur in sequential order. Using these assumptions, it was found that the expected numbers of neurons having b. b - 1, b - 2.….1 input error bits in their state update are equal. Because of the capability of error correction, an asynchronous HAM is also found to have a higher Pdc than a synchronous HAM. Using our results, network reliability and generalization capability can be estimated when both the interconnection faults and the number of error bits in the probe vectors are taken into account.
Original language | English |
---|---|
Pages (from-to) | 969-980 |
Number of pages | 12 |
Journal | IEEE Transactions on Neural Networks |
Volume | 3 |
Issue number | 6 |
DOIs | |
Publication status | Published - 1992 Nov |
All Science Journal Classification (ASJC) codes
- Software
- Computer Science Applications
- Computer Networks and Communications
- Artificial Intelligence