Systolic Simulation of Hamming Neural Network
Jacek Mazurkiewicz · 2003
The paper describes the systolic simulation of Hamming net algorithm. Systolic arrays creation methodology is based on Data Dependence Graphs analysis. The proposed implementation is focused on completely digital circuits. Input data is passed through the neurons in a time share basis, weights are stored in digital shift registers and no separate thresholds are used. The architecture provides massive parallelism and reprogrammability. The efficiency of proposed simulation is discussed based on implementation quality criteria for array architecture.