A Comprehensive Review and a Taxonomy of Edge Machine Learning: Requirements, Paradigms, and Techniques

By Wenbin Li et al

Published on Aug. 15, 2023

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Edge Machine Learning: Requirements
3. ML Requirements
4. EC Requirements
5. Overall Requirements
6. Summary of Edge ML Techniques
7. Analysis of Techniques
8. Open Issues and Future Directions
9. Conclusion

Summary

The document provides a comprehensive review and taxonomy of Edge Machine Learning, focusing on the requirements, paradigms, and techniques involved. It discusses the union of Edge Computing and Artificial Intelligence, emphasizing the challenges and opportunities in the edge environment. The paper outlines the computational and environmental constraints for ML on the edge, detailing ML requirements such as low task latency, high performance, generalization, adaptation, privacy, and security. It also covers EC requirements including computational efficiency, optimized bandwidth, and offline capability. The overall requirements encompass availability, reliability, energy efficiency, and cost optimization. Through an exhaustive review of over twenty Edge ML techniques, the paper aims to fill a significant gap in the literature by offering a comprehensive view of the landscape of Edge ML.