Dynamical constraints on using precise spike timing to compute in recurrent cortical networks.

Area of Science:

• Computational neuroscience
• Neural dynamics
• Spiking neural networks

Background:

• Models of cortical computation increasingly utilize precise spike timing.
• Experimental data show neurons in thalamus and sensory cortices exhibit high temporal precision.
• Spike-timing-based models require dynamics insensitive to initial conditions.

Purpose of the Study:

• To investigate the feasibility of robust precise spike trajectory generation in recurrent cortical networks.
• To identify criteria for network dynamics sensitivity to initial conditions.
• To analyze the dynamics of specific recurrent cortical architectures.

Main Methods:

• Development of an abstract dynamical system for spiking neural networks.
• Identification of a criterion for stationary dynamics sensitivity to initial conditions.
• Analysis of recurrent cortical architectures, including those in orientation selectivity research.

Main Results:

• Under sustained, low to moderate, weakly correlated cortical activity, recurrent networks are unlikely to generate precise spike trajectories.
• The criterion for sensitivity to initial conditions was applied to analyze network dynamics.
• Analysis revealed limitations in robust millisecond-timescale spatiotemporal pattern generation.

Conclusions:

• Recurrent cortical networks likely cannot reliably produce millisecond-precise spike timing under typical physiological conditions.
• The findings challenge the widespread assumption of precise spike timing as a primary mechanism for cortical computation.
• Further research may explore alternative computational principles or network configurations.