Neuromorphic Implementation of Attractor Dynamics in a Two-Variable Winner-Take-All Circuit with NMDARs: A Simulation...

Area of Science:

• Neuroscience
• Neuromorphic Engineering
• Computational Neuroscience

Background:

• Winner-take-all (WTA) neural networks with N-methyl-D-aspartate receptor (NMDAR)-mediated synaptic dynamics exhibit attractor characteristics crucial for cognitive functions.
• Implementing complex cognitive functions like decision-making, working memory, and hysteresis in neuromorphic systems remains a challenge.

Purpose of the Study:

• To present a novel method for neuromorphic implementation of a two-variable WTA circuit with NMDARs.
• To enable the circuit to perform decision-making, working memory, and hysteresis in visual perception tasks.
• To provide a biophysically plausible model for understanding cognitive dynamics.

Main Methods:

• A dynamical system approach to circuit synthesis was employed, based on a biophysically plausible WTA model.
• The circuit was designed to generate slow and non-linear temporal dynamics characteristic of NMDAR-mediated synapses.
• Circuit simulations were performed using Cadence, and theoretical analysis of the dynamical system was conducted.

Main Results:

• Circuit simulations successfully reproduced ramping neural activities during decision-making tasks.
• Sustained neural activities, akin to those in the prefrontal cortex during working memory, were observed.
• Classical hysteresis behavior was replicated in simulations of visual discrimination tasks.
• Theoretical analysis elucidated the mechanisms underlying decision-making, memory capacity, and hysteresis loops.

Conclusions:

• The developed WTA circuit with NMDARs effectively captures attractor dynamics essential for cognitive functions.
• The consistency between circuit simulations and theoretical analysis validates the proposed neuromorphic implementation.
• These circuits serve as promising elementary modules for building integrated neuromorphic cognitive systems.