Abstract
The dynamics of a convergent iterative unlearning algorithm proposed earlier [7, 8] is examined. A self-consistent system of equations of the spectral dynamics of a synaptic matrix is obtained at the thermodynamic limit. The unlearning intensity (which varies during the iteration process) that optimizes the algorithm's rate of convergence on the projector matrix is found. The synaptic-matrix spectrum dynamics for optimal unlearning is determined.
Similar content being viewed by others
References
J. J. Hopfield, Proc. Natl. Acad. Sci. USA,79, No. 8, 2554 (1982).
D. J. Amit, H. Gutfreund, and H. Sompolinsky, Ann. Phys.,173, No. 1, 30 (1987).
S. Diederich and M. Opper, Phys. Rev. Lett.,58, No. 9, 949 (1987).
W. Krauth and M. Mezard, J. Phys. A,20, No. 11, L745 (1987).
M. G. Blatt and E. G. Vergini, Phys. Rev. Lett.,66, No. 13, 1793 (1991).
K. W. Berryman, M. E. Inchiosa, A. M. Jaffe, and S. A. Janowsky, J. Phys. A,23, No. 5, L223 (1990).
A. Yu. Plakhov and S. A. Semenov, Neural Network World,4, 405 (1993).
A. Yu. Plakhov and S. A. Semenov, J. Phys. I (France), 4, No. 2 (1994).
L. Personnaz, I. Guyon, and G. Dreyfus, J. Phys. (France) Lett.,46, No. 8, L359 (1985).
I. Kanter and H. Sompolinsky, Phys. Rev. A,35, No. 1, 380 (1987).
V. A. Marchenko and L. A. Pastur, Matem. Sb.,72(114), No. 4, 507 (1967).
M. Opper, Europhys. Lett.,8, No. 4, 389 (1989).
Additional information
Institute of Physicotechnical Problems, Moscow. Translated from Izvestiya Vysshikh Uchebnykh Zavedenii, Radiofizika, Vol. 37, No. 9, pp. 1104–1115, September, 1994.
Rights and permissions
About this article
Cite this article
Plakhov, A.Y. Unlearning dynamics in Hopfield neural network. Radiophys Quantum Electron 37, 711–718 (1994). https://doi.org/10.1007/BF01039610
Received:
Issue Date:
DOI: https://doi.org/10.1007/BF01039610