Activity Stabilization in a Population Model of Working Memory by Sinusoidal and Noisy Inputs

Area of Science:

• Computational Neuroscience
• Cognitive Neuroscience
• Neural Dynamics

Background:

• Working memory (WM) theories propose information retention via persistent neural spiking.
• The relationship between WM activity and observed oscillatory profiles is not well understood.
• Investigating the roles of gamma-band oscillations and noise in WM dynamics is crucial.

Purpose of the Study:

• To explore the joint effects of gamma-band oscillations and noise on WM models.
• To understand how these inputs stabilize the metastable active regime in neural networks.
• To elucidate the impact of coupling speed and oscillation phase on WM stabilization.

Main Methods:

• Utilized firing rate models of working memory, starting with a single excitatory-inhibitory circuit.
• Extended analysis to coupled systems of two circuits and a multi-circuit system with clusters.
• Investigated the influence of common vs. independent noise and in-phase vs. anti-phase gamma inputs.

Main Results:

• Both gamma-band oscillations and noise can stabilize the active regime supporting WM retention.
• Fast coupling enhances stabilization by common noise, amplified by in-phase gamma inputs.
• Gamma-band input differentially stabilizes common-noise groups, with stronger effects for synchronized gamma and fast connections.

Conclusions:

• Local gamma oscillations can stabilize neural activity in large-scale WM representations.
• Fast long-range connections and synchronized gamma oscillations amplify the stabilizing effect.
• Findings provide mechanistic insights into how neural oscillations and noise support working memory.