A Unified Granular Fuzzy-Neuro Min-Max Relational Learner: A Case Study

This paper deals with a real world problem of medical diagnosis, to this goal, we propose to learn a compact fuzzy medical knowledge base through a cognitively-motivated granular hybrid neuro-fuzzy or fuzzy-neuro possibilistic model appropriately crafted as a means to automatically extract fuzzy weighted production rules. The main idea is to start learning from coarse fuzzy partitions of the involved proteins variations of input variables and proceed progressively toward fine-grained partitions until finding the appropriate partitions that fit the data. We provide details of implementation issues, experimental results, and discussion of interpretability issues. Moreover, learning is firmly grounded on fuzzy relational calculus, linguistic approximation and the crucial notion of importance widely used in human decision making and clinical problem-solving.