Ex) Article Title, Author, Keywords
New Phys.: Sae Mulli 2020; 70: 547-555
Published online June 30, 2020 https://doi.org/10.3938/NPSM.70.547
Copyright © New Physics: Sae Mulli.
Myoung Won CHO*
Department of Global Medical Science, Sungshin Women's University, Seoul 01133, Korea
This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
The implementation of neural network content-addressable memory (CAM) relates to the attractor dynamics in the firing process.
If some firing patterns are carved as fixed-point attractors, one of them can be recalled, depending on initial firing states.
Many studies following the landmark achievement by Hopfield, have suggested an algorithm based on an analogy between the asymptotic dynamics of neural networks and the equilibrium properties of magnetic systems. However, the firing process of biological neurons, which progresses depending on minute spiking timings, has distinctive properties of attractor dynamics with those of the classical neuron models. We here study the characteristics of fixed-point attractors in a spiking neural network model preserving the attribute of spiking-timing-dependent interactions between neurons. We show that the attractor dynamics in the model relate to the phase-locking dynamics in a biological neural network when the memorized firing states are orthogonal to one another. We also introduce how non-orthogonal firing states can be memorized in the model and how one of them can be recalled, depending on the initial states, by applying the Hopfield network.
Keywords: Neural network learning and dynamics, Spiking neural network, Content addressable memory