Research Progress and Development of Deep Learning Based on Convolutional Neural Network

Abstract

As a popular branch of the field of artificial intelligence and machine learning, deep learning has received increasing attention and continuous development. This paper first introduces the development and current situation of deep learning. Then, it introduces the component and mathematical theories of convolutional neural network (CNN). As for CNN optional variables and parameters, the optimal range of each parameter tested is explored through the training and tests on the datasets of Fashion-MNIST and CIFAR20 respectively. Finally, this paper proposes existing defects and future development of deep learning based on CNN.