Neural-network model composed of multidimensional spin neurons
Authors:
- 1 February 1995
- journal article
- research article
- Published by American Physical Society (APS) in Physical Review E
- Vol. 51 (2) , 1538-1546
Abstract
As an extension of the Hopfield model, we propose a neural network composed of D-dimensional spin neurons (D≥1). Our model is equivalent to the Hopfield model in the case of D=1 and is related to the clock neural network in the case of D=2. We derive the free energy of our model using the replica symmetric theory. When a finite number of patterns are embedded they are found to be retrievable if the tempeature T is lower than 1/D. The phase diagram and the storage capacity of the network are also obtained with the storage capacity =0.0743 (D=2) and =0.0432 (D=3) for T=0.
Keywords
This publication has 13 references indexed in Scilit:
- Capacity of diluted multi-state neural networksJournal of Physics A: General Physics, 1994
- Retrieval and chaos in extremely dilutedQ-Ising neural networksJournal of Statistical Physics, 1994
- Stability properties of Potts neural networks with biased patterns and low loadingJournal of Physics A: General Physics, 1991
- Modeling Brain Function1989
- Discrete-state phasor neural networksPhysical Review A, 1988
- Associative Memory in Sparse Phasor Neural NetworksEurophysics Letters, 1988
- Potts-glass models of neural networksPhysical Review A, 1988
- Statistical mechanics of neural networks near saturationAnnals of Physics, 1987
- Storing Infinite Numbers of Patterns in a Spin-Glass Model of Neural NetworksPhysical Review Letters, 1985
- Spin-glass models of neural networksPhysical Review A, 1985