Design of clustering analyzer based on systolic array architecture
Mao‐Fu Lai, Yan‐Pei Wu, Chaur‐Heh Hsieh · 2002
This paper presents a systolic architecture for the squared-error clustering algorithm. The proposed architecture exploits a 2-dimensional systolic array which uses intensively parallel and pipelined processing. The architecture dramatically reduces the huge number of processing elements required by previous architectures. Furthermore, the same organization can be utilized for applications where the number of input patterns is varied. In addition, the time complexity of our architecture is reduced in comparison with earlier architectures. A cost-effective VLSI implementation for high speed clustering analysis can be realized with considerably less circuit complexity using this novel architecture.