[Paper Review] A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects

Zewen Li, Wenjie Yang+2 more|arXiv (Cornell University)|Apr 1, 2020

Advanced Neural Network Applications84 references264 citations

TL;DR

This paper surveys Convolutional Neural Networks (CNNs) from history to current state, covering 1D, 2D, and multi-dimensional CNNs, with experimental insights and future directions.

ABSTRACT

Convolutional Neural Network (CNN) is one of the most significant networks in the deep learning field. Since CNN made impressive achievements in many areas, including but not limited to computer vision and natural language processing, it attracted much attention both of industry and academia in the past few years. The existing reviews mainly focus on the applications of CNN in different scenarios without considering CNN from a general perspective, and some novel ideas proposed recently are not covered. In this review, we aim to provide novel ideas and prospects in this fast-growing field as much as possible. Besides, not only two-dimensional convolution but also one-dimensional and multi-dimensional ones are involved. First, this review starts with a brief introduction to the history of CNN. Second, we provide an overview of CNN. Third, classic and advanced CNN models are introduced, especially those key points making them reach state-of-the-art results. Fourth, through experimental analysis, we draw some conclusions and provide several rules of thumb for function selection. Fifth, the applications of one-dimensional, two-dimensional, and multi-dimensional convolution are covered. Finally, some open issues and promising directions for CNN are discussed to serve as guidelines for future work.

Motivation & Objective

• Provide a broad overview of CNN history and developments.
• Summarize classic and advanced CNN models and their key innovations.
• Offer experimental insights and practical guidelines for function selection.
• Discuss applications across 1D, 2D, and multi-dimensional convolutions.
• Identify open issues and promising directions for future research.

Proposed method

• Synthesize historical context and CNN overview.
• Review classic and advanced CNN architectures and their pivotal innovations.
• Analyze experimental results to derive conclusions and rules of thumb for function selection.
• Discuss applications spanning 1D, 2D, and multi-dimensional convolutions.
• Highlight open issues and propose future research directions.

Experimental results

Research questions

• RQ1 What are the key milestones and developments in CNN history and model design?
• RQ2 What empirical conclusions and practical rules of thumb can be drawn from CNN experiments?
• RQ3 How are 1D, 2D, and multi-dimensional convolutions applied across domains?
• RQ4 What open issues and promising directions exist for CNN research?

Key findings

• The survey aims to provide novel ideas and prospects for CNNs in a fast-growing field.
• It covers one-dimensional, two-dimensional, and multi-dimensional convolutions.
• It offers an overview of CNN history, models, and key points that enable state-of-the-art results.
• Experimental analysis leads to conclusions and practical rules of thumb for function selection.
• The paper discusses applications and identifies open issues and future directions.