Does the Cerebral Cortex Exploit High-Dimensional, Non-linear Dynamics for Information Processing?

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Stimulus-induced synchronization in the visual cortex suggests temporal relationships in neuronal discharges encode feature relations.
• Temporal coherence was proposed as a signature of perceptual grouping, prompting extensive research.
• Previous studies confirmed some predictions but revealed complex dynamics beyond simple coherence, whose role remains unclear.

Purpose of the Study:

• To present a framework linking various cortical dynamics.
• To assign coding functions to low-dimensional (e.g., synchronized oscillations, phase shifts) and high-dimensional (non-linear, non-stationary) dynamics.
• To propose a unified view of cortical signal processing.

Main Methods:

• Utilized chronic multisite recordings from the visual cortex of anesthetized cats.
• Analyzed data from monkeys performing cognitive tasks.
• Developed a synthetic framework to connect different dynamical manifestations.

Main Results:

• Established links between low-dimensional dynamics (synchronized oscillations, large-scale correlations) and high-dimensional dynamics.
• Identified low-dimensional dynamics as substates representing computational results.
• Demonstrated that these substates arise from computations within high-dimensional state-spaces of recurrently coupled networks.

Conclusions:

• Cortical dynamics involve both low-dimensional and high-dimensional features with distinct coding functions.
• Synchronized oscillations and large-scale correlations represent computational outcomes.
• These dynamics are embedded within the complex state-space of recurrent neural networks.