A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects

This review surveys convolutional neural networks (CNNs), one of the most significant model families in deep learning, from a general perspective rather than focusing narrowly on individual application scenarios as many earlier reviews did. It aims to incorporate recently proposed ideas and covers not only 2-D convolution but also 1-D and multidimensional variants, walking through the history of CNNs, an overview of different convolution types, and both classic and advanced CNN models along with the key design choices that let them reach state-of-the-art results.

Through experimental analysis, the authors draw conclusions and offer several rules of thumb for selecting functions and hyperparameters, then survey applications of 1-D, 2-D, and multidimensional convolution. The review closes by discussing open issues and promising directions, serving as a broad guideline to orient future work in this fast-growing field.