On interpretability of fuzzy models

Abstract

Interpretability is one of the indispensable features of fuzzy models. This paper discusses the interpretability of fuzzy models with/without prior knowledge about the target system. Without prior knowledge, conciseness of fuzzy models helps humans to interpret their input-output relationships. In the case where a human has the knowledge in advance, an interpretable model could be the one that explicitly explains his/her knowledge. Experimental results show that the concise model has the essential interpretable feature. The results also show that human’s knowledge changes the most interpretable model from the most concise model.