Networked Artificial Intelligence: AI-Enabled 5G Networking

The integration of fifth generation (5G) wireless technologies with distributed artificial intelligence (AI) is transforming network operations. AI is increasingly embedded in all network elements, from cloud and edge to terminal devices, enabling AI to function as a networking system. This convergence facilitates AI-based applications across the global network, with notable successes in various domains such as computer vision, natural language processing, and healthcare. Networked Artificial Intelligence: AI-Enabled 5G Networking a comprehensive framework for the deep integration of computing and communications, optimizing networks and applications as a unified system using AI. The book covers topics ranging from networked AI fundamentals to AI-enabled 5G networks, including agent modeling, machine learning (ML) algorithms, and network protocol architectures. It discusses how network service providers can leverage AI and ML techniques to customize network baselines, reduce noise, and accurately identify issues. It also looks at AI-driven networks that enable self-correction for maximum uptime and prescriptive actions for issue resolution, as well as troubleshooting by capturing and storing data before network events. The book presents a comprehensive approach to AI-enabled networking that offers unprecedented opportunities for efficiency, reliability, and innovation in telecommunications. It works through the approach’s five steps of connection, communication, collaboration, curation, and community. These steps enhance network effects, empowering operators with insights for trusted automation, cost reduction, and optimal user experiences. The book also discusses AI and ML capabilities that enable networks to continuously learn, self-optimize, and predict and rectify service degradations proactively, even with full automation. Cover Half Title Title Copyright Dedication Contents Preface Author Chapter 1 Networked Artificial Intelligence 1.1 Emergence of AI Technology 1.1.1 Machine Learning 1.1.2 Deep Learning and Neural Networks 1.1.3 Generative Artificial Intelligence 1.2 Usage of Artificial Intelligence 1.3 Artificial Intelligence Enabling and Supporting Technologies 1.3.1 Natural Language Processing 1.3.2 Computer Vision 1.3.3 Graphical Processing Units 1.3.4 Internet-of-Things 1.3.5 Advanced Algorithms 1.3.6 Application Programming Interfaces 1.4 Emergence of Extremely High-Speed Semiconductor Technologies 1.5 Emergence of Extremely High-Bandwidth Networking Technologies 1.5.1 Terrestrial Wireline Networks 1.5.2 Fifth- and Sixth-Generation Wireless Networks 1.5.3 High-Bandwidth Satellite Networks 1.6 Open Systems Interconnection 1.7 Internet Protocol Layer Model 1.8 Summary Chapter 2 Artificial Intelligent Agent 2.1 Agent and Environment 2.2 Agent System 2.2.1 Controller 2.2.2 Hierarchical Controller 2.2.3 Body 2.2.4 Environment 2.3 Agent Architecture and Control 2.4 Summary Chapter 3 Agent Function 3.1 Time 3.2 Tracing Percept 3.3 Tracing Command 3.4 History 3.5 Memory/Belief State 3.6 Rationality 3.7 Online and Offline Learning 3.8 Action 3.8.1 Solution Searching 3.8.2 Reasoning 3.8.3 Planning 3.8.4 Action with Certainty 3.8.5 Action with Uncertainty 3.8.6 Action in Learning 3.9 Summary Chapter 4 Agent Modeling 4.1 Agent 4.2 Reasoning 4.2.1 Cognitive Reasoning 4.2.2 Reasoning with Constraints 4.2.3 Reasoning with Uncertainty 4.2.4 Knowledge Reasoning 4.3 Learning 4.4 Learning with Reasoning 4.5 Federated Learning 4.6 Ensemble Learning 4.7 Machine Learning with Logical Reasoning 4.7.1 Hybrid Learning 4.8 Summary Chapter 5 Multi-Agent System 5.1 Overview 5.2 Collaborative Prognostics 5.3 Agent Typologies and Failure Modes 5.4 Value Asset 5.5 Digital Twin 5.6 Mediator Agents 5.7 Social Platform 5.8 Multi-Agent Architecture 5.8.1 Centralized 5.8.2 Hierarchical 5.8.3 Heterarchical 5.8.4 Distributed 5.9 Summary Chapter 6 Protocol Layer Architecture 6.1 Open Standard International Model 6.1.1 Layer 1: Physical Layer 6.1.2 Layer 2: Data Link Layer 6.1.3 Layer 3: Network Layer 6.1.4 Layer 4: Transport Layer 6.1.5 Layer 5: Session Layer 6.1.6 Layer 6: Presentation Layer 6.1.7 Layer 7: Application Layer 6.2 Internet Model 6.2.1 Layer 1: Physical Layer 6.2.2 Layer 2: Data Link Layer 6.2.3 Layer 3: Internet/Network Layer 6.2.4 Layer 4: Transport Layer 6.2.5 Layer 5–6: Middleware 6.2.6 Layer 7: Application Layer 6.3 Peer-to-Peer Architecture Model 6.4 Summary Chapter 7 Artificial Intelligence Performance Analysis 7.1 Signal/Data Acquisition 7.2 Signal/Data Digitization 7.3 Signal/Data Clearing and Preprocessing 7.4 Feature Extraction 7.5 Dimensionality Reduction 7.6 Signal/Data Classification Techniques 7.7 Signal/Data Classification Evaluation Criteria 7.7.1 Confusion Matrix 7.7.2 Macro- and Micro-Averaging 7.7.3 Fβ-Score 7.7.4 Matthews Correlation Coefficient (MCC) 7.7.5 Receiver Operating Characteristics (ROC) 7.7.6 Area under the ROC Curve (AUC) 7.7.7 Examples 7.7.8 Use of ROC Curve 7.8 Summary Chapter 8 Unsupervised Machine Learning 8.1 Hierarchical Clustering 8.2 Bayesian Clustering 8.3 Partitional Clustering 8.3.1 K-Means Clustering 8.3.2 Mixture Models 8.4 Applications of Clustering in Networks 8.5 Latent Variable Model 8.5.1 Mixture Distribution 8.5.2 Factor Analysis 8.5.3 Blind Signal Separation 8.5.4 Non-Negative Matrix Factorization 8.5.5 Hidden Markov Models 8.5.6 Bayesian Networks and Probabilistic Graph Models 8.5.7 Applications of Latent Variable Models in Networks 8.6 Dimensionality Reduction 8.6.1 ISOMAP 8.6.2 Generative Topographic Mapping 8.6.3 Locally Linear Embedding 8.6.4 Principal Curve 8.6.5 Nonlinear Multi-Dimensional Scaling 8.6.6 T-Distributed Stochastic Neighbor Embedding 8.7 Outlier Detection 8.7.1 Nearest Neighbor Anomaly Detection 8.7.2 Local Outlier Factor 8.7.3 Connectivity-Based Outlier Factors 8.7.4 Influenced Outlierness 8.7.5 Local Outlier Probability Models 8.8 Key Characteristic of Unsupervised Learning 8.9 Applications of Unsupervised Learning in Networking 8.10 Future Network: Research Challenges and Opportunities 8.10.1 Simplified Network Management 8.10.2 Semi-Supervised Learning for Computer Networks 8.10.3 Transfer Learning in Computer Networks 8.10.4 Federated Learning in Computer Networks 8.10.5 Generative Adversarial Networks in Computer Networks 8.11 Pitfalls and Caveats of Using ML/DL in Networking 8.11.1 Inappropriate Technique Selection 8.11.2 Lack of Interoperability of Some Supervised ML/DL Algorithms 8.11.3 Lack of Operations Success of ML/DL Networking 8.11.4 Ignoring Simple Non-ML/DL Based Tools 8.11.5 Overfitting 8.11.6 Data Quality Issues 8.11.7 Inaccurate Model Building 8.11.8 Machine Learning in Adversarial Environments 8.12 Summary Chapter 9 Supervised Machine Learning 9.1 Linear Regression and Classification 9.2 Logistic Regression 9.3 Ridge Regression 9.4 LASSO Regression 9.5 Tree-Based Models 9.5.1 Decision Tree 9.5.2 Random Forests 9.5.3 Gradient Boosting Regression 9.5.4 XGBoost 9.5.5 Light Gradient Boosted Machine Regressor 9.6 Summary Chapter 10 Deep Learning 10.1 Cost and Error Function, Gradient Decent, and Backpropagation in Neural Network 10.1.1 Gradient Descent in Neural Networks 10.1.2 Two-Dimensional Gradient Descent Example and Backpropagation Depth 10.1.3 Vectorization in Neural Networks 10.1.4 Matrix Multiplication 10.1.5 Cost/Error/Loss Function Derivatives for Gradient Descent 10.1.6 Propagating into Hidden Layers 10.1.7 Vectorization of Backpropagation 10.1.8 Implementing Gradient Descent Step 10.1.9 Final Gradient Descent Algorithm 10.2 Summary Chapter 11 Overfitting and Underfitting 11.1 Overfitting 11.2 Resolving Overfitting 11.2.1 Linear Regression 11.2.2 Ridge Regression 11.2.3 Lasso Regression 11.2.4 Early Stopping 11.2.5 Cross-Validation 11.2.6 Train with More Data 11.2.7 Data Augmentation 11.2.8 Feature Selection 11.3 Underfitting 11.3.1 Resolving Underfitting 11.4 Summary Chapter 12 Hybrid Learning 12.1 Semi-Supervised Learning 12.2 Self-Supervised Learning (SSL) 12.3 Multi-Instance Learning 12.4 Contrastive Self-Supervised Learning 12.5 Non-Contrastive Self-Supervised Learning 12.6 Self-Supervised Semi-Supervised Learning (S4L) 12.7 Summary Chapter 13 Reinforcement Learning 13.1 RL Agent and Environment Interaction 13.1.1 Policies 13.2 Optimal Stationary Policy for Infinite Horizon Problems 13.3 Value of a Policy 13.4 Q-Table 13.5 Value of an Optimal Policy 13.6 Policy Categories 13.7 Temporal Difference 13.7.1 Temporal Differences (TD) 13.8 Policy Gradient 13.9 Expected Return 13.10 V-Value Function 13.11 Q-Value Function 13.12 Fitted Q-Learning 13.13 Deep Q-Networks 13.14 Massive Parallel DQN Architecture 13.15 DDQN: Double Deep Q-Networks 13.15.1 Background 13.15.2 Deep Q Networks (DQN) 13.15.3 Double Deep Q-Networks (Double DQN) 13.16 Neural Fitted Q-Network (NFQ) 13.17 Advantage Actor-Critic Network 13.17.1 Estimating Qπ 13.17.2 Estimating Vπ 13.18 Asynchronous Advantage Actor-Critic Network 13.18.1 Asynchronous Advantage Actor-Critic 13.19 Dueling Deep Q-Network 13.20 Summary Chapter 14 Artificial Intelligence Application and Network Protocol Architecture Model 14.1 Artificial Intelligence-Aware Applications Services 14.2 Artificial Intelligence-Enabled/Standard-Based Applications 14.3 Artificial Intelligence-Enabled/Standard-Based Middleware Infrastructure 14.3.1 Artificial Intelligence-Standard-Based Middleware 14.3.2 Artificial Intelligence-Enabled Middleware 14.4 Artificial Intelligence-Enabled/Standard-Based Transport Protocols 14.5 Artificial Intelligence-Enabled/Standard-Based Network and Routing Protocols 14.6 Artificial Intelligence-Enabled/Standard-Based Link/Medium Access Control (MAC) Protocols 14.7 Artificial Intelligence-Enabled/Standard-Based Physical (PHY) Layer 14.8 Summary Chapter 15 AI-Enabled Network 15.1 AI-Enabled Physical Layer 15.2 AI-Enabled Link/Medium Access Control (MAC) Layer 15.3 AI-Enabled Network Layer 15.4 AI-Enabled Transport Network Layer 15.5 AI-Enabled Middleware Layer 15.6 AI-Enabled Session Layer 15.7 AI-Enabled Presentation Layer 15.8 AI-Enabled Application Layer 15.8.1 AI/ML/DL-Enabled Cybersecurity 15.9 Summary Chapter 16 AI-Enabled End-to-End Network 16.1 Overview 16.2 Artificial Intelligence and Multidisciplinary Applications 16.3 Common AI/ML/DL Communications Network Infrastructure 16.4 AI/ML/DL Networking Architecture 16.5 End-to-End AI-Enabled OSI Layer Using Common AI/ML/DL Infrastructure 16.6 Networked AI-Enabled Applications 16.7 Summary Chapter 17 AI-Enabled Peer-to-Peer Network 17.1 Summary Chapter 18 Artificial Intelligence-Enabled 5G Network 18.1 Overview 18.2 5G Radio Access Network 18.3 5G Radio Interface 18.4 5G Core Network 18.5 5G Protocol Stack 18.5.1 5G Radio Network Nodes vs 5G Deployment Types 18.5.2 5G User and Control Planes for eNB, gNB, and ng-eNB 18.5.3 5G Control Plane: UE-to-AMF and UE-to-SMF Protocol Stack 18.5.4 5G User Plane: UE-to-AMF and UE-to-SMF Protocol Stack 18.5.5 5G Non-Standalone versus 5G Standalone Architecture 18.6 5G Security 18.6.1 5G Non-Standalone NR Security 18.6.2 Evolution of the 5G Trust Model 18.6.3 5G Phase 1 Security (Release 15) 18.7 5G Network Slicing 18.8 5G MAC Protocol 18.9 5G Higher Layer Protocols 18.10 AI-Enabled 5G Applications 18.11 5G End-to-End Network Architecture with AI-Enabled Applications 18.12 Summary Index The integration of fifth-generation (5G) wireless technologies with distributed artificial intelligence (AI) is transforming network operations. AI is increasingly embedded in all network elements, from cloud and edge to terminal devices, enabling AI to function as a networking system. This convergence facilitates AI-based applications across the global network, with notable successes in various domains such as computer vision, natural language processing, and healthcare. Networked Artificial Intelligence: AI-Enabled 5G Networking is a comprehensive framework for the deep integration of computing and communications, optimizing networks and applications as a unified system using AI. The book covers topics ranging from networked AI fundamentals to AI-enabled 5G networks, including agent modeling, machine learning (ML) algorithms, and network protocol architectures. It discusses how network service providers can leverage AI and ML techniques to customize network baselines, reduce noise, and accurately identify issues. Also, it looks at AI-driven networks that enable self-correction for maximum uptime and prescriptive actions for issue resolution, as well as troubleshooting by capturing and storing data before network events. The book presents a comprehensive approach to AI-enabled networking that offers unprecedented opportunities for efficiency, reliability, and innovation in telecommunications. It works through the approach’s five steps of connection, communication, collaboration, curation, and community. These steps enhance network effects, empowering operators with insights for trusted automation, cost reduction, and optimal user experiences. The book also discusses AI and ML capabilities that enable networks to continuously learn, self-optimize, and predict and rectify service degradations proactively, even with full automation.