ISSN 0869-6632 (Print)
ISSN 2542-1905 (Online)

For citation:

Soloviev I. A., Klinshov V. V. Stability thresholds of attractors of the Hopfield network. Izvestiya VUZ. Applied Nonlinear Dynamics, 2023, vol. 31, iss. 1, pp. 75-85. DOI: 10.18500/0869-6632-003028, EDN: CFEFMZ

This is an open access article distributed under the terms of Creative Commons Attribution 4.0 International License (CC-BY 4.0).
Full text PDF(Ru):
PDF icon and_2023-1_soloviev-klinshov_75-85.pdf
(downloads: 0)
Full text PDF(En):
PDF icon and_2023_1_soloviev_klinshov_eng.pdf
(downloads: 232)
Language: 
Russian
Heading: 
Applied Problems of Nonlinear Oscillation and Wave Theory
Article type: 
Article
UDC: 
530.182
DOI: 
10.18500/0869-6632-003028
EDN: 
CFEFMZ

Stability thresholds of attractors of the Hopfield network

Autors: 
Soloviev Igor Aleksandrovich, Institute of Applied Physics of the Russian Academy of Sciences
Klinshov Vladimir Viktorovich, Lobachevsky State University of Nizhny Novgorod
Abstract: 

Purpose of the work is the detailed study of the attractors of the Hopfield network and their basins of attraction depending on the parameters of the system, the size of the network and the number of stored images. To characterize the basins of attraction we used the method of the so-called stability threshold, i.e., the minimum distance from an attractor to the boundary of its basin of attraction. For useful attractors, this value corresponds to the minimum distortion of the stored image, after which the system is unable to recognize it. In the result of the study it is shown that the dependence of the average stability threshold of useful attractors on the number of stored images can be nonmonotonic, due to which the stability of the network can improve when new images are memorized. An analysis of the stability thresholds allowed to estimate the maximum number of images that the network can store without fatal errors in their recognition. In this case, the stability threshold of useful attractors turns out to be close to the minimum possible value, that is, to unity. To conclude, calculation of the stability thresholds provides important information about the attraction basins of the network attractors.

Key words: 
dynamical networks
collective dynamics
associative memory
Acknowledgments: 
This work was supported by Russian Foundation for Basic Research, grant No 20-52-12021
Reference: 
  1. Hopfield JJ. Neural networks and physical systems with emergent collective computational abilities. Proc. Natl. Acad. Sci. U. S. A. 1982;79(8):2554–2558. DOI: 10.1073/pnas.79.8.2554.
  2. Hopfield JJ. Neurons with graded response have collective computational properties like those of two-state neurons. Proc. Natl. Acad. Sci. U. S. A. 1984;81(10):3088–3092. DOI: 10.1073/pnas. 81.10.3088.
  3. Farhat NH, Psaltis D, Prata A, Paek E. Optical implementation of the Hopfield model. Applied Optics. 1985;24(10):1469–1475. DOI: 10.1364/AO.24.001469.
  4. Hoppensteadt FC, Izhikevich EM. Pattern recognition via synchronization in phase-locked loop neural networks. IEEE Transactions on Neural Networks. 2000;11(3):734–738. DOI: 10.1109/ 72.846744.
  5. Joya G, Atencia MA, Sandoval F. Hopfield neural networks for optimization: study of the different dynamics. Neurocomputing. 2002;43(1–4):219–237. DOI: 10.1016/S0925-2312(01)00337-X.
  6. Wen UP, Lan KM, Shih HS. A review of Hopfield neural networks for solving mathematical programming problems. European Journal of Operational Research. 2009;198(3):675–687. DOI: 10.1016/j.ejor.2008.11.002.
  7. McEliece R, Posner E, Rodemich E, Venkatesh S. The capacity of the Hopfield associative memory. IEEE Transactions on Information Theory. 1987;33(4):461–482. DOI: 10.1109/TIT.1987.1057328.
  8. Storkey A. Increasing the capacity of a hopfield network without sacrificing functionality. In: Gerstner W, Germond A, Hasler M, Nicoud JD. editors. Artificial Neural Networks — ICANN’97. ICANN 1997. Vol. 1327 of Lecture Notes in Computer Science. Berlin, Heidelberg: Springer; 1997. P. 451–456. DOI: 10.1007/BFb0020196.
  9. Krotov D, Hopfield JJ. Dense associative memory for pattern recognition. In: NIPS’16: Proceedings of the 30th International Conference on Neural Information Processing Systems. 5–10 December 2016, Barcelona, Spain. New York: Curran Associates Inc.; 2016. P. 1180–1188. DOI: 10.5555/ 3157096.3157228.
  10. Ramsauer H, Schafl B, Lehner J, Seidl P, Widrich M, Adler T, Gruber L, Holzleitner M, Pavlovic M, Sandve GK, Greiff V, Kreil D, Kopp M, Klambauer G, Brandstetter J, Hochreiter S. Hopfield networks is all you need [Electronic resource]. arXiv:2008.02217. arXiv Preprint; 2020. 94 p. Available from: https://arxiv.org/abs/2008.02217.
  11. Klinshov VV, Nekorkin VI, Kurths J. Stability threshold approach for complex dynamical systems. New Journal of Physics. 2016;18(1):013004. DOI: 10.1088/1367-2630/18/1/013004.
  12. Menck PJ, Heitzig J, Marwan N, Kurths J. How basin stability complements the linear-stability paradigm. Nature Physics. 2013;9(2):89–92. DOI: 10.1038/nphys2516.
  13. Chakraborty A, Alam M, Dey V, Chattopadhyay A, Mukhopadhyay D. Adversarial attacks and defences: A survey [Electronic resource]. arXiv:1810.00069. arXiv Preprint; 2018. 31 p. Available from: https://arxiv.org/abs/1810.00069.
  14. Amari SI, Maginu K. Statistical neurodynamics of associative memory. Neural Networks. 1988;1(1):63–73. DOI: 10.1016/0893-6080(88)90022-6.
Received: 
24.11.2022
Accepted: 
26.12.2022
Available online: 
18.01.2023
Published: 
31.01.2023
  • 1518 reads