Abstract
First, we study the effects of introducing training noise on the retrieval behaviours of dilute attractor neural networks. We found that, in general, training noise enhances associativity, but also reduces the attractor overlap. At a narrow range of storage levels, however, the system exhibits re-entrant retrieval behaviour on increasing training noise.
Secondly, we consider optimization of network performance, and subsequently the storage capacity, in the presence of retrieval noise (temperature). This is achieved by adapting the network to an appropriate training overlap, which is determined self-consistently by the optimal attractor overlap. The maximum storage capacity deviates from the storage capacity of the maximally stable network on increasing temperature, and in the high temperature regime (T ≥ 0.38 for Gaussian noise, the Hebb-rule network yields the maximum storage capacity. Our analysis demonstrates the principles of specialization and adaptation in neural networks.
Preview
Unable to display preview. Download preview PDF.
References
Gardner E and Derrida B 1988 J Phys A: Math Gen 21 271
Wong K Y M and Sherrington D 1988 Europhys Lett 7 197
Wong K Y M and Sherrington D 1990 J Phys A: Math Gen 23 L175
Wong K Y M and Sherrington D 1990 “Optimally adapted attractor neural networks in the presence of noise” submitted to J Phys A: Math Gen
Derrida B, Gardner E and Zippelius A 1987 Europhys Lett 4 167
Amit D, Gutfreund H and Sompolinsky H 1987 Ann Phys (NY) 173 30
Amit D, Evans M, Horner H and Wong K Y M 1990 J Phys A: Math Gen in press
Gardner E, Stroud N and Wallace D 1989 J Phys A: Math Gen 22 2019
Rosenblatt F 1962 Principles of Neurodynamics (New York: Spartan)
Minsky M and Papert S 1969 Perceptrons (Cambridge, MA: MIT Press)
Gardner E 1988 J Phys A: Math Gen 21 257
Kepler T B and Abbott L F 1988 J Physique 49 1657
Krauth W, Nadal J-P and Mezard M 1988 J Phys A: Math Gen 21 2995
Gardner E 1989 J Phys A: Math Gen 22 1969
Krauth W and Mezard M 1987 J Phys A: Math Gen 22 2019
Wong K Y M and Sherrington D 1990 “Retrieval properties of noise-optimal attractor neural networks” to be published
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1990 Springer-Verlag
About this paper
Cite this paper
Wong, K.Y.M., Sherrington, D. (1990). Tailoring the performance of attractor neural networks. In: Garrido, L. (eds) Statistical Mechanics of Neural Networks. Lecture Notes in Physics, vol 368. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3540532676_44
Download citation
DOI: https://doi.org/10.1007/3540532676_44
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-53267-5
Online ISBN: 978-3-540-46808-0
eBook Packages: Springer Book Archive