Operating in a Reverberating Regime Enables Rapid Tuning of Network States to Task Requirements
Published in Frontiers in Systems Neuroscience, 2018
Abstract:
Neural networks need to process information on different timescales, which requires a delicate balance between stability and flexibility. Recent theoretical work suggests that this balance can be achieved by operating in a “reverberating regime” where activity persists for an intermediate duration. In this regime, networks can rapidly adapt to changing task requirements while maintaining stable representations. We show that this regime is characterized by a specific relationship between excitation and inhibition, and that small changes in network parameters can shift the system between different dynamical states. Our results provide insights into how neural networks can efficiently process information across multiple timescales.
Recommended citation: Wilting, J., Dehning, J., Pinheiro Neto, J., Rudelt, L., Wibral, M., Zierenberg, J., & Priesemann, V. (2018). Operating in a Reverberating Regime Enables Rapid Tuning of Network States to Task Requirements. Frontiers in Systems Neuroscience, 12(November).
Download Paper