Keywords
associative memory
Hebbian learning rule
recognition
How to Cite
Abstract
this study examines the properties of a fully connected neural network composed of phase neurons, following the Hebbian learning rule. The signals transmitted through the network’s interconnections are single pulses with specific phases. A neuron’s firing rule is defined as follows: among the total signals received at a neuron’s input, the phase component with the highest amplitude is identified, and the neuron emits a single pulse with the same phase. The phases encoding the components of associative memory vectors are randomly distributed. To estimate the recognition error, we employ the Chernov-Chebyshev technique, which is independent of the phase encoding distribution type. Our findings demonstrate that the associative memory capacity of this neural network is four times greater than that of a traditional Hopfield network that operates with binary patterns. Consequently, the radius of the attraction region is also four times larger.
References
Hopfield J. Neural Networks and Physical Systems with Emergent Collective Computational Abilities. Proc. Nat. Acad. Sci. USA. 1982;79:2554–2558.
Hebb D. The Organization of Behavior. New York: Wiley; 1949. 365 р.
Palm G., Sommer F. Information Capacity in Recurrent McCulloch-Pitts Networks with Sparsely Coded Memory States. Network. 1992;3:177–186.
Kuh A., Dickinson B. Information Capacity of Associative Memory. IEEE Trans. on Inf. Theory. 1989;35:59–68.
Herz A., Marcus C. Distributed Dynamics in Neural Networks. Phys. Rev. E. 1993;47:2155–2161.
McEllice R., Posner E. et al. Capacity of the Hopfield Associative Memory. IEEE Trans. on Inf. Theory. 1987;33:461–482.
Bolle D., Dupont P., Mourik J. van. Stability Properties of Potts Neural Networks with Biased Pattern and Low Loading. J. Phys. A. 1991;24:1065.
Kazanovich Y., Borisyuk R. Dynamics of Neural Networks with Central Element. Neural Networks. 1999;12:441–454.
Kryzhanovsky B. V., Magomedov B. M., Mikaelian A. L. A Domaine Model of Neural Network. Doklady Mathematics. 2005;71:310–314.
Vogt H., Zippelius A. Invariant Recognition in Potts-Glass Neural Network. J. Phys. A. 1992;25:2209.
Sompolinsky H. Neural Network with Non-linear Synapses and Static Noise. Phys. Rev. A. 1986;34:2571.
Крыжановский Б. В., Микаэлян А. Л. О распознающей способности нейросети на нейронах с параметрическим преобразованием частот. Доклады АН. Сер. Мат. физика. 2002;383(3):318–321.
Крыжановский Б. В., Микаэлян А. Л. Ассоциативная память, способная распознавать сильно скоррелированные образы. Доклады АН. Сер. Информатика. 2003;390(1):27–31.
Mikaelian A. L., Kryzhanovsky B. V., Litinskii L. B. Parametrical Neural Network. Optical Memory&Neural Network. 2003;12(3):227–236.
Kryzhanovsky B. V., Litinskii L. B., Mikaelian A. L. Vector-Neuron Models of Associative Memory. 2004 IEEE International Joint Conference on Neural Networks. Budapest; 2004. Р. 909–1004.
Kanter I. Potts-Glass Models of Neural Networks. Phys. Rev. A. 1988;37(7):2739.
Bolle D., Dupont P., Huyghebaert J. Thermodynamic Properties of the q-State Potts-Glass Neural Network. Phys. Rev. A. 1992;45:4194.
Wu F. Y. The Potts Model. Rev. Mod. Phys. 1982;54:235.
Amit D. J., Gutfreund H., Sompolinsky H. Information Storage in Neural Networks with Low Levels of Activity. Phys. Rev. A. 1987;35:2293.
Karandashev I., Kryzhanovsky B., Litinskii L. Weighted Patterns as a Tool to Improve the Hopfield Model. Physical Review E. 2012;85:041925.