A Comparative Study between Artificial Neural Networks and Optical Neural Networks

Artificial intelligence has gained widespread interest from people from all walks of life since the big data era arrived, so artificial neural networks (ANNs) are one of the most important areas of artificial intelligence that have a close relationship and an effective role in diverse artificial intelligence applications, data is loaded between memory and processor during the calculation, which reduces the calculation rate and increases the power consumption. Training artificial neural network parameters take a long time and consume a lot of energy. Therefore, it is necessary to use new neural networks that differ from artificial neural networks, called optical neural networks. Optical neural networks can overcome the above shortcomings. Compared with artificial neural networks, they have obvious advantages such as high speed, high bandwidth, high balanced processing, low latency, and low power consumption. The optical neural network can significantly improve the performance of required tasks by taking advantage of the high-speed characteristics. The study compares artificial neural networks and optical neural networks to train networks by nonlinear activation functions to perform image classification tasks. The Neural networks are trained and tested with the help of the KMNIST dataset. The backward propagation occurs using the backpropagation algorithm.