A Measure of Synergy Based on Union Information

Area of Science:

• Information Theory
• Computational Neuroscience
• Machine Learning

Background:

• Partial Information Decomposition (PID) aims to partition information from multiple sources about a target into unique, redundant, and synergistic components.
• Classical information theory lacks a unique framework for this decomposition, necessitating additional assumptions.
• Union information, representing information present in at least one source, offers a path to synergy measurement.

Purpose of the Study:

• Introduce a novel measure of union information from a communication channel perspective.
• Compare the new measure with existing union and synergy measures.
• Critically review existing characterizations of union and synergy in PID.

Main Methods:

• Developed a new union information measure grounded in communication channel principles.
• Performed comparative analysis against established PID measures.
• Conducted a comprehensive literature review of union and synergy characterizations.

Main Results:

• The proposed union information measure provides a new avenue for PID.
• The study elucidates the properties and comparisons of this novel measure.
• Identified gaps and areas for improvement in current PID literature.

Conclusions:

• The novel communication channel-based union information measure offers a promising approach to PID.
• This work contributes to a more rigorous understanding of information synergy and redundancy.
• Further research is warranted to explore the full implications of this new measure in various scientific domains.