A Review of Structure Optimization of Convolutional Neural Networks

Abstract

In recent years, Convolutional neural network (CNNs) has gained a significant development in the industrial manufacturing process. However, its application in the environment with high real-time requirements and limited resources is restricted by its huge scale and complicated computing degree. The optimization of the structure of convolutional neural networks has become a research hotspot in the field of deep learning. In this paper, development history, research status and typical methods of structural optimization technology of convolutional neural networks are summarized, which are summarized into five aspects of pruning & sparsification, tensor factorization, knowledge transferring, compacting module designing and automatic design. And a more comprehensive discussion is carried out. Finally, the developments and difficult points of current research in this paper will be analyzed and summarized, and the future development direction and application prospect of network structure optimization field also will be forecast.