Utilizing optical neural network to establish high-performance OR and XOR logic gates

The optical-neural-network logic gates using unsupervised learning method and supervised learning method are investigated. The structures of the optical neurons using self-connection configuration and interconnection configuration are proposed. The performance of the AND, OR, NAND, NOR and XOR logic gates are analyzed. According to our simulation results, the bit error ratio (BER) of the optical neurons using the interconnection configuration is lower than that using self-connection configuration. For OR logic gate, the best performance is BER = 6.54%. For XOR logic gate, the best performance is BER < 4.89 x 10-5. The results show that the proposed optical structure can work for different logic gates by tuning the parameters of the couplers and the phase shifters.