Fuzzy Measure Parametrization from Densities in Information Fusion

A new paper on arXiv (2607.23243) addresses the challenge of parametrizing Fuzzy Measures (FM) for aggregation in information fusion, such as ensemble methods and decision-level fusion. The study shows that density information alone—weights of individual sources—is insufficient to uniquely determine a discrete FM, though an interval-valued FM can be uniquely identified. The authors demonstrate that incorporating additional information, such as the choice of a specific Fuzzy Integral and a dataset, enables the determination of a unique FM. This work highlights limitations of widely used approaches like Sugeno-λ and Decomposable FMs, which extrapolate from densities while respecting monotonicity constraints.