Negation-Aware Weighted Information Fusion for Reliable Evidential Reasoning

Abstract:Dempster--Shafer theory (DST) provides an attractive framework for multi-source information fusion. However, its effectiveness critically depends on how a BPA is constructed for each source. Existing methods often rely on heuristic mappings or restrict belief assignment to singleton hypotheses and the full set, thereby underutilizing the expressive belief structure of BPAs for representing this http URL paper proposes Negation-Aware Weighted Information Fusion (NAWIF).First, NAWIF transforms the class-wise possibility distribution of each source into a set of negation-aware simple BPAs, which encode class exclusion through complementary propositions and retain residual uncertainty in the full set. Their combination produces a closed-form BPA over composite hypotheses, preserving ambiguity among the classes not excluded by the source. It further performs source-specific BPA discounting based on its reliability and uncertainty, suppressing the impact of unreliable or uninformative this http URL on benchmark pattern recognition datasets demonstrate the effectiveness of NAWIF, which achieves state-of-the-art average performance with particularly pronounced advantages under noise disturbance.

Machine Learning (cs.LG)

arXiv:2509.00754 [cs.LG]

(or arXiv:2509.00754v3 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2509.00754 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.