A current-feedback method for programming memristor array in bidirectional associative memory

Memristor-based realization of artificial neural network (ANN) circuits require efficient and accurate methods for programming memristance which represents the synaptic weight. A novel feedback-based method is proposed in this paper for programming memristor array, which trades off the time-consuming of programming and circuit complexity. A case of size 6×6 bidirectional associative memory (BAM) neural network is introduced for verification of the proposed method in Cadence integrated circuit simulation environment using Verilog-AMS. The weight matrix of BAM is learned by software MATLAB with a training set of 3 pairs of Tetris patterns. Simulation result shows that the correctness of the programmed memristive BAM circuit and the effectiveness of the proposed method that can be adopted for setting synapse weight in other ANN circuit designs.