product iamge

Social Edge Computing: Empowering Human-Centric Edge Computing, Learning and Intelligence

This document was uploaded by one of our users. The uploader already confirmed that they had the permission to publish it. If you are author/publisher or own the copyright of this documents, please report to us by using this DMCA report form.

Download
Search on Amazon

Simply click on the Download Book button.

Yes, Book downloads on Ebookily are 100% Free.

Sometimes the book is free on Amazon As well, so go ahead and hit "Search on Amazon"

The rise of the Internet of Things (IoT) and Artificial Intelligence (AI) leads to the emergence of edge computing systems that push the training and deployment of AI models to the edge of networks for reduced bandwidth cost, improved responsiveness, and better privacy protection, allowing for the ubiquitous AI that can happen anywhere and anytime. Motivated by the above trend, this book introduces a new computing paradigm, the Social Edge Computing (SEC), that empowers human-centric edge intelligent applications by revolutionizing the computing, intelligence, and the training of the AI models at the edge. The SEC paradigm introduces a set of critical human-centric challenges such as the rational nature of edge device owners, pronounced heterogeneity of the edge devices, real-time AI at the edge, human and AI interaction, and the privacy of the edge users. The book addresses these challenges by presenting a series of principled models and systems that enable the confluence of the computing capabilities of devices and the domain knowledge of the people, while explicitly addressing the unique concerns and constraints from humans.

 

Compared to existing books in the field of edge computing, the vision of this book is unique: we focus on the social edge computing (SEC), an emerging paradigm at the intersection of edge computing, AI, and social computing. This book discusses the unique vision, challenges and applications in SEC. To our knowledge, keeping humans in the loop of edge intelligence has not been systematically reviewed and studied in an existing book. The SEC vision generalizes the current machine-to-machine interactions in edge computing (e.g., mobile edge computing literature), and machine-to-AI interactions (e.g., edge intelligence literature) into a holistic human-machine-AI ecosystem. 

Author(s): Dong Wang, Daniel 'Yue' Zhang
Publisher: Springer
Year: 2023

Language: English
Pages: 183
City: Cham

Preface
Acknowledgments
Contents
1 A New Human-Centric Computing Age at Edge
1.1 Social Edge: A Human-Centric Definition of ``Edge''
1.1.1 Human-in-the-Loop Computing at the Edge
1.1.2 Human-in-the-Loop Intelligence at the Edge
1.1.3 Human-in-the-Loop Learning at the Edge
1.2 Human-Centric Design: A Double-Edged Sword
1.2.1 The Rational Edge Challenge
1.2.2 The Pronounced Heterogeneity Challenge
1.2.3 The Human-AI Integration Challenge
1.2.4 The Human Responsiveness and Quality Assurance Challenge
1.2.5 The Privacy and Performance Trade-Off Challenge
1.2.6 The Crowd Data Imbalance Challenge
1.3 Contributions and Organization
References
2 Social Edge Trends and Applications
2.1 Social Edge Computing: The Paradigm Shift
2.1.1 What Is Social Edge Computing?
2.1.1.1 Human-Centric Nature of SEC
2.1.1.2 Flexibility of SEC to Support System Variations
2.1.2 Why We Need Social Edge Computing?
2.2 Enabling Technologies for Social Edge Computing
2.2.1 Edge Computing
2.2.2 Social Sensing
2.2.3 Edge AI
2.2.4 Federated Learning
2.3 Emerging Social Edge Computing Applications
2.3.1 Disaster and Emergency Response
2.3.2 Collaborative Traffic Monitoring
2.3.3 Crowd Abnormal Event Detection
2.3.4 Automatic License Plate Recognition
2.3.5 Crowd Video Sharing
References
3 Rational Social Edge Computing
3.1 The Rational Social Edge Computing Problem
3.2 A Cooperative-Competitive Game-theoretic Tasks Allocation Framework
3.2.1 Problem Definition
3.2.2 The CoGTA Framework
3.2.2.1 Cooperative-Competitive Game for Task Allocation
3.2.2.2 Game Protocol and Payoff Function
3.2.2.3 Decentralized Fictitious Play with Negotiation
3.2.2.4 Dynamic Incentive Adjustment
3.3 Real-World Case Studies
3.4 Discussion
References
4 Taming Heterogeneity in Social Edge Computing
4.1 The Heterogeneous Social Edge
4.2 A Heterogeneous Social Edge System: HeteroEdge
4.2.1 Problem Definition
4.2.2 The HeteroEdge Framework
4.2.2.1 Runtime and Hardware Abstraction
4.2.2.2 Supply Chain-Based Resource Management
4.3 Real-World Case Studies
4.4 Discussion
References
5 Real-Time AI in Social Edge
5.1 The Real-Time Social Edge
5.2 A Real-Time Optimal Edge Batching System: EdgeBatch
5.2.1 Problem Definition
5.2.2 The EdgeBatch Framework
5.2.2.1 Stochastic Optimal Task Batching Module (SOTB)
5.2.2.2 Optimal Contracting with Asymmetric Information
5.3 Real-World Case Studies
5.4 Discussion
References
6 Human-AI Interaction
6.1 Interactive Human-Machine Edge Intelligence
6.2 A Crowd-AI Hybrid System: CrowdLearn
6.2.1 Problem Definition
6.2.2 The CrowdLearn Framework
6.2.2.1 Query Set Selection (QSS)
6.2.2.2 Incentive Policy Design (IPD)
6.2.2.3 Crowd Quality Control (CQC)
6.2.2.4 Machine Intelligence Calibration (MIC)
6.3 Real-World Case Studies of CrowdLearn
6.4 Incorporating Human Visual Attention
6.4.1 The iDSA Framework
6.4.1.1 Crowd Task Generation to Acquire Human Perception
6.4.1.2 Budget Constrained Adaptive Incentive Policy
6.4.1.3 Interactive Attention Convolutional Neural Network
6.4.1.4 Social Media Image Normalization
6.5 Real-World Case Studies of iDSA
6.6 Discussion
References
7 Privacy in Social Edge
7.1 Understanding Privacy in Social Edge
7.2 A Privacy-Aware Framework for Distributed Edge Learning: FedSens
7.2.1 Problem Definition
7.2.2 Framework Overview and Model Intuition
7.2.3 Device Selection with Intrinsic-Extrinsic Deep Reinforcement Learning
7.2.4 Adaptive Global Update Control
7.2.5 Summary of FedSens Workflow
7.3 Real-World Case Studies
7.4 Discussion
References
8 Further Readings
8.1 Social Sensing
8.2 Edge Computing
8.3 Task Allocation in Real-Time Systems
8.4 Distributed System with Heterogeneous Computing Nodes
8.5 IoT Middleware
8.6 Human-AI Systems and Active Learning
8.7 Learning with Imbalanced Data
References
9 Conclusion and Remaining Challenges
9.1 Conclusion and Summary
9.2 Remaining Challenges
9.2.1 Security Against Malicious Crowd
9.2.2 Robustness Against Churn and Dynamic Context
9.2.3 Uncertainty Quantification in SEC
9.2.4 Blurring Human-Machine Boundaries with Social Edge Graph
References
Index