Evidence Combination From an Evolutionary Game Theory Perspective.

Area of Science:

• Artificial Intelligence
• Information Fusion
• Computational Biology

Background:

• Dempster-Shafer evidence theory is crucial for multisource information fusion, effectively handling uncertain data.
• The standard Dempster's rule of combination can yield counter-intuitive outcomes in certain scenarios.
• Existing methods to mitigate these issues often focus on minimizing information loss or deviation.

Purpose of the Study:

• To propose a novel evolutionary combination rule (ECR) inspired by evolutionary game theory.
• To address counter-intuitive results in Dempster-Shafer evidence theory from a biological and evolutionary standpoint.
• To enhance the synthesis of multiple pieces of evidence in uncertain information systems.

Main Methods:

• Developed an evolutionary combination rule (ECR) based on biological and evolutionary principles.
• Formalized proposition interactions using a Jaccard matrix game.
• Employed replicator dynamics to simulate the evolution of propositions within the ECR framework.

Main Results:

• The proposed ECR effectively suppresses counter-intuitive behaviors observed in evidence theory paradoxes.
• ECR demonstrates superior performance compared to numerous existing methods in experimental evaluations.
• Mathematical proofs confirm the stability and convergence properties of the ECR solutions.

Conclusions:

• The ECR offers a biologically supported approach to multisource information fusion.
• This method provides a robust alternative for synthesizing uncertain information, overcoming limitations of traditional rules.
• The study highlights the potential of evolutionary game theory in advancing evidence combination methodologies.