Working memory is the process through which information is held in mind temporarily. To achieve this neural representations of the data must persist while the information is being held in mind. However, these representations cannot be held in mind indefinitely. Instead, they must be flexible enough to switch between different pieces of information. Here we propose a computational model of a network of excitatory and inhibitory neurons that is capable of generating persistent but also flexible representations of information. We find that by operating near criticality this model can maximize persistence and flexibility. In this model, unlike prevailing models of criticality in neural systems, criticality is a dynamic regime on the edge between a winner-take-all phase and an asynchronous phase. At criticality, dynamics recordings were used to confirm the multifaceted predictions from the model. This confirmation suggests that in the rat prefrontal cortex, the neural circuits operate near criticality while performing working memory tasks.
Contact: Jacob Barfield, barfieldjh@hollins.edu
Hollins University, Roanoke, VA
Additional Authors:
Woodrow L. Shew