Abstract
In this paper we analyze the pattern storage capacity of the exponential correlation associative memory (ECAM). This architecture was first studied by Chiueh and Goodman [3] who concluded that, under certain conditions on the input patterns, the memory has a storage capacity that was exponential in the length of the bit-patterns. A recent analysis by Pelillo and Hancock [9], using the Kanerva picture of recall, concluded that the storage capacity was limited by 2N−1/N 2 patterns. Both of these analyses can be criticised on the basis that they overlook the role of initial bit-errors in the recall process and deal only with the capacity for perfect pattern recall. In other words, they fail to model the effect of presenting corrupted patterns to the memory. This can be expected to lead to a more pessimistic limit. Here we model the performance of the ECAM when presented with corrupted input patterns. Our model leads to an expression for the storage capacity of the ECAM both in terms of the length of the bit-patterns and the probability of bit-corruption in the original input patterns. These storage capacities agree closely with simulation. In addition, our results show that slightly superior performance can be obtained by selecting an optimal value of the exponential constant.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
Hopfield, J. J.: Neural networks and physical systems with emergent collective com-putational abilities, Proc. Natl. Acad. Sci. USA 79 (1982), 2554–2558.
McEliece, R. J., Posner, E. C., Rodemich, E. R., and Venkatesh, S. S.: The capacity of the Hopfield associative memory, IEEE Trans. Inform. Theory 33 (1987), 461–482.
Chiueh, T. D. and Goodman, R. M.: Recurrent correlation associative memories, IEEE Trans. Neural Networks 2 (1991), 275–284.
Chiueh, T. D. and Goodman, R. M.: VLSI implementation of a high-capacity neural network associative memory, In D. S. Touretzky, (ed.), Advances in Neural Information Processing Systems 2, Morgan-Kaufmann, San Mateo, CA, (1990), pp. 793–800.
Wang, C. C. and Don, H. S.: An analysis of high-capacity discrete exponential BAM, IEEE Trans. Neural Networks 6 (1995), 492–496.
Wang C. C. and Lee, J. P.: The decision-making properties of discrete multiple exponential bidirectional associative memories, IEEE Trans. Neural Networks 6 (1995), 993–999.
Chiueh, T. D. and Tsai, H. K.: Multi-valued associative memories based on recurrent networks. IEEE Trans. Neural Networks 4 (1993), 364–366.
Hancock, E. R. and Kittler, J.: Discrete Relaxation, Pattern Recognition 23 (1990), 711–733.
Hancock, E. R. and Pelillo, M.: A Bayesian Interpretation for the Exponential Corre-lation Associative Memory, Pattern Recognition Letters 19 (1998), 149–159.
Wilson, R. C. and Hancock, E. R.: Structural Matching by Discrete Relaxation, IEEE Trans. Pattern Anal. Machine Intell. 19 634–648, 1997.
Kanerva, P.: Sparse Distributed Memory, MIT Press, Cambridge, MA (1988).
Milun, D. and Sher, D.: Improving sampled probability distributions for Markov ran-dom fields, Pattern Recognition Letters 14 (1993), 781–788.
Chou, P. A.: The capacity of the Kanerva associative memory is exponential, in D. Z. Anderson (ed.), Neural Information Processing Systems, American Institute of Physics, New York, NY (1988), pp. 184–191.
Hassoun, M. and Watta, P.: The Hamming Associative Memory and its Relation to the Exponential Correlation DAM, IEEE International Conference on Neural Networks 1 (1996), 583–587.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Wilson, R.C., Hancock, E.R. Storage Capacity of the Exponential Correlation Associative Memory. Neural Processing Letters 13, 71–80 (2001). https://doi.org/10.1023/A:1009603617772
Issue Date:
DOI: https://doi.org/10.1023/A:1009603617772