Frontiers in Computational Neuroscience (Feb 2013)

Stability analysis of associative memory network composed of stochastic neurons and dynamic synapses

  • Yuichi eKatori,
  • Yuichi eKatori,
  • Yosuke eOtsubo,
  • Masato eOkada,
  • Masato eOkada,
  • Kazuyuki eAihara

DOI
https://doi.org/10.3389/fncom.2013.00006
Journal volume & issue
Vol. 7

Abstract

Read online

We investigate the dynamical properties of an associative memory network consisting of stochastic neurons and dynamic synapses that show short-term depression and facilitation. In the stochastic neuron model used in this study, the efficacy of the synaptic transmission changes according to the short-term depression or facilitation mechanism. We derive a macroscopic mean field model that captures the overall dynamical properties of the stochastic model. We analyze the stability and bifurcation structure of the mean field model, and show the dependence of the memory retrieval performance on the noise intensity and parameters that determine the properties of the dynamic synapses, i.e., time constants for depressing and facilitating processes. The associative memory network exhibits a variety of dynamical states, including the memory and pseudo-memory states, as well as oscillatory states among memory patterns. This study provides comprehensive insight into the dynamical properties of the associative memory network with dynamic synapses.

Keywords