Revealing the Dynamics of Neural Information Processing with Multivariate Information Decomposition

Area of Science:

• Neuroscience
• Computational Neuroscience
• Information Theory

Background:

• Neural circuits process information, a fundamental question in neuroscience.
• Multivariate information theory offers new frameworks for analyzing neural computation.
• Classical information theory provides a basis for understanding information processing.

Purpose of the Study:

• To introduce the partial information decomposition (PID) framework for analyzing neural information processing.
• To explore conceptual and practical issues in applying PID to neural circuits.
• To discuss recent empirical work utilizing PID in neuroscience.

Main Methods:

• Introduction to the partial information decomposition (PID) framework.
• Analysis of redundant, unique, and synergistic information integration modes.
• Focus on synergistic information, representing higher-order input patterns.

Main Results:

• Synergistic information dynamics are widespread in neural circuitry.
• These dynamics show structure-function relationships.
• Synergistic information emerges in specific neural structures like rich clubs and recurrent networks.

Conclusions:

• Partial information decomposition (PID) provides valuable insights into neural information processing.
• Synergistic information is a key component of complex neural computations.
• Future directions include applying PID in behaving animals and time-resolved analyses.