Learning Fuzzy Measures

© 2020, Springer Nature Switzerland AG. This chapter is a key contribution of this work in which various computational approaches to learning fuzzy measures are described. The learning problem is framed from the perspective of data fitting, where we aim to define a model that interpolates or approximates a set of observed or user-specified instances. Fitting is performed with respect to different metrics, and by solving different convex and non-convex optimisation problems. The computational complexity of fuzzy measures is addressed by using simplifying assumptions, in particular the notion of k-order fuzzy measures and the software packages implementing the presented fitting approaches are also described.