Deep Learning Applications on Edge Computing

Abstract

In various applications, such as computer vision and natural language processing, deep learning is commonly used. End devices like smartphones and sensors on the Internet of Things (IoT) are used to provide data to be analyzed with deep learning or to train profound models in real time. Deep learning and teaching, however, require considerable computational resources. Edge calculation is a feasible method of fulfilling high computation and low-latency needs for deep learning on edge equipment and it offers additional benefits in data security, bandwidth efficiency, and scalability. This chapter aims to provide a thorough overview of current state-of-the-art technologies of the computer sciences’ crossroads. It specifically provides an overview of applications in which deep learning is used at the network level; describes various approaches for the fast execution of deep learning inference across a combination of final devices, edge servers, and the cloud; and describe training models for multi-level devices. It discusses open issues in systems performance, network and management technology, benchmarks, and privacy. The reader will rule out the following concepts from the article: an understanding of the scenarios for profound learning at the network edge, an understanding popular technological techniques to accelerate profound learning lessons and distributed training on cutting-edge devices, and the knowledge of new developments and opportunities.