A system design methodology for fuzzy clustering neural networks

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