Fuzzy clustering neural network system design and implementation

In this paper, a novel VLSI system design methodology for fuzzy clustering neural networks (FCNN) is presented. This methodology emphasizes a coordination between model definition, architectural description, and hardware implementation, Two mapping strategies, from FCNN model to parallel architecture and from architecture to systolic implementation, are discussed. The system design is achieved by 1) developing an effective FCNN model, where a direct fuzzy competitive learning algorithm between the nodes is adopted; 2) designing the corresponding parallel architecture with special feedforward and feedback paths; 3) building the systolic array (SA) suitable for VLSI.