Some Insights Into Convolutional Neural Networks

As a kind of deep learning model, convolutional neural networks (CNNs) have greatly boosted the state-of-the-art performance and have found their successful applications in many fields, such as computer version, pattern recognition, natural language processing, etc. Many distinguished CNN models, for example, AlexNet, Google inception net, VGGNet, and so on, have been developed for various tasks. In this paper, we provide some insights into convolutional neural networks in three aspects: activation function, convolution operation, pooling operation, and we also discuss the training of CNNs. The insights presented in this paper will provide the potential readers with further understanding of CNNs.