Automatic fuzzy encoding of complex objects

In this work we propose an approach to encode real objects represented by feature vectors into fuzzy concepts. A system is designed, whose main component is an adaptive fuzzy encoder which learns object membership degrees to fuzzy concepts from a set of objects and the expert's crisp assignment to a concept. To properly use the expert's crisp choices, the learning is performed with the help of a fuzzy decoder that translates the membership values provided by the fuzzy encoder into crisp information. Moreover, a mapping is defined to improve the interpretability of the knowledge acquired through learning by the fuzzy encoder.