Self-Organizing Gaussian Fuzzy CMAC with Truth Value Restriction

The cerebellar model articulation controller (CMAC) is a popular auto-associate memory feed forward neural network model. Since it was proposed, many researchers have introduced fuzzy logic to CMAC and called FCMAC. In FCMAC, the input data is fuzzificated into fuzzy sets before fed into CMAC. This paper proposes self-organizing fuzzification (SOF) technique to form fuzzy sets in the fuzzification phase. The proposed SOF technique uses raw numerical values of a training data set with no preprocessing and obtains dynamic partition-base clusters without prior knowledge of number of clusters. It also provides CMAC a consistent fuzzy rule base. Truth value restriction inference scheme (TVR) is employed in the defuzzification phase. Our experiments are conducted on some benchmark datasets, and the results show that our method outperforms the existing model with higher ability to handle uncertainty in the inference process