Rule Simplification

Abstract

The interpretability of FRBs might be hampered by the existence of a large number of rules or complicated ones. Thus, any attempt to use the FARB–ANN equivalence for knowledge extraction from large-scale networks must include a systematic approach for rule reduction and simplification.

Simplification procedures usually lead to a more interpretable rule-base with the cost of a degradation in the FRB performance. This tradeoff between complexity and interpretability can be viewed as one manifestation of Zadeh’s principle of incompatibility (see Section 1.2).