Inductive learning using similarity measures on lattice-fuzzy set

This paper is concerned with learning concept description from fuzzy data. A concept is defined by a set of examples and counter-examples. We propose an inductive learning algorithm that finds fuzzy rules that recognize almost all of the examples and almost none of the counter-examples. Contrary to usual representation of fuzziness, lattice fuzzy sets are used to modelize uncertainty and imprecision. A case-based approach of the learning process is proposed, based on similarity measures defined on lattice structures. An application on sponge data illustrates the interest of the learning algorithm proposed