Decomposing multivariate information rates in networks of random processes.

Area of Science:

• Complex Systems
• Information Theory
• Network Science

Background:

• Partial Information Decomposition (PID) analyzes interdependencies in complex networks.
• Traditional PID assumes memorylessness, limiting its use in dynamic systems with temporal correlations.

Purpose of the Study:

• To introduce Partial Information Rate Decomposition (PIRD), extending PID to dynamic processes.
• To decompose dynamic information in multivariate random processes into unique, redundant, and synergistic contributions.

Main Methods:

• Leveraging mutual information rate (MIR) instead of mutual information (MI).
• Defining a pointwise redundancy rate function using the minimum MI principle in the frequency domain.
• Validating the framework with Gaussian systems and physiological network data.

Main Results:

• PIRD effectively captures temporal correlations missed by traditional PID.
• Spectral representation reveals scale-specific higher-order interactions.
• Frequency-dependent redundant information exchange was observed in cerebrovascular and cardiovascular variables.

Conclusions:

• PIRD provides a principled way to decompose information in dynamic systems.
• Accounting for temporal structure and spectral content is crucial for information decomposition in neuroscience and physiology.
• PIRD offers a powerful tool for analyzing dynamic network behavior.