深層 SIRMs 結合型ファジィ推論モデルにおける未知の知識獲得に関する一考察

抄録

Deep learning, which has been the focus of much attention in recent years, provides highly accurate results, but has the problem of unclear input-output relationships. On the other hand, fuzzy inference models that use If-Then rules can represent human knowledge and make the inference process easy to understand. One such model is the deep SIRMs coupled fuzzy inference model. This model is characterized by its single-input rules, which not only make the rules easy to understand, but also realize exclusive OR. However, rules that use the output of the previous layer as one of the input variables of the next layer are difficult to understand.
In this study, we propose a new interpretation of the output of each layer of the deep SIRMs coupled fuzzy inference model to discover new conditional attributes that are useful for inference. We also evaluate the feasibility of the method using medical diagnosis data.