FCMAC-Yager: a novel Yager-inference-scheme-based fuzzy CMAC

Area of Science:

• Neuroscience and Artificial Intelligence
• Fuzzy Logic and Neural Networks

Background:

• The cerebellum is crucial for motor control, cognitive functions, and learning.
• The Cerebellar Model Articulation Controller (CMAC) neural network mimics cerebellar functions but has interpretability and quantization issues.
• Existing CMAC limitations hinder its application in complex, real-world scenarios.

Purpose of the Study:

• To address CMAC limitations by integrating fuzzy logic and advanced clustering techniques.
• To develop a novel fuzzy neural architecture, the FCMAC-Yager system, combining CMAC's learning with fuzzy inference's reasoning.
• To evaluate the FCMAC-Yager system's performance in diverse real-life applications.

Main Methods:

• Developed the Fuzzy CMAC (FCMAC) by applying fuzzy quantization to the CMAC structure.
• Integrated the Yager fuzzy inference scheme into the FCMAC, creating the FCMAC-Yager system.
• Employed a two-phase training algorithm: Discrete Incremental Clustering (DIC) for structural learning and Hebbian learning for parameter learning.

Main Results:

• The FCMAC-Yager system successfully integrates cerebellar-like learning and memory with human-like fuzzy reasoning.
• Demonstrated effective application in highway traffic-trend modeling, bank failure prediction, and breast cancer diagnosis.
• Experimental results indicate encouraging performance and potential for early warning systems.

Conclusions:

• The FCMAC-Yager system offers a robust and interpretable fuzzy neural architecture.
• This novel system overcomes CMAC's traditional limitations, enhancing its applicability.
• The FCMAC-Yager system shows significant promise for complex predictive modeling and classification tasks.