Systolic architectures for high order correlation artificial neural nets
Weiquan Bao, Magdy Bayoumi · 2003
Systolic architectures for high-order correlation neural nets are proposed. They are based on using a multiply associated high-order correlation tensor as a mathematical model. The case of triple-order correlation nets is analyzed. A design procedure based on developing a triple-order correlation dependence graph is presented. The developed method is flexible. Several implementation issues are discussed.>