Similarity in hierarchical fuzzy rule - base systems

This paper deals with issues related to fuzzy interpolation and hierarchical fuzzy rule bases, from the point of view of similarity between rule antecedents and observations. The existing distance-based similarities are exhibiting sometimes strange properties, which make them inadequate for the selection of interpolating fuzzy rules, for instance. In the paper, similarity of two fuzzy sets is expressed as a crisp value, rather then based on a fuzzy distance. This leads to easier rule selection, but also to some modification of the interpolation methods.