Efficient two-dimensional systolic array architecturefor multilayer neural network
Hesham H. Amin, K.M. Curtis, Barrie Hayes‐Gill · Electronics Letters · 1997
An efficient two-dimensional systolic array architecture for a multilayer perceptron neural network is presented. This structure can perform the weighted sum calculation in only N cycles, for an N × N network. A new technique is introduced to eliminate the problem of long connecting lines in the proposed systolic structure.