Convergence of Artificial Intelligence and the Internet of Things (eBook)

Introduction 6
Research Solutions 7
Conclusion 13
References 13
Contents 15
Fog Computing: Data Analytics for Time-Sensitive Applications 17
1 Introduction 17
2 Fog Computing Applications 18
3 Architecture of Fog Computing 20
3.1 Smart Layer 20
3.2 Fog Layer 21
3.3 Cloud Services 21
4 Benefits of Fog Computing 21
5 Challenges of Fog Computing 23
6 Conclusion and Discussions 26
References 27
Medical Image Watermarking in Four Levels Decomposition of DWT Using Multiple Wavelets in IoT Emergence 30
1 Introduction 31
2 Digital Image Watermarking Algorithms 32
2.1 Biorthogonal Wavelet 33
2.2 Reverse Biorthogonal Wavelet 33
2.3 Symlet Wavelet 34
2.4 Coiflets Wavelet 34
2.5 Discrete Meyer Wavelet 35
3 The Proposed Medical Image Watermarking Algorithm 37
4 Experimental Results and Evaluation 40
5 Conclusion 44
References 45
Optimised Statistical Model Updates in Distributed Intelligence Environments 47
1 Introduction 47
1.1 Problem Statement 48
1.2 Paper Organisation 48
2 Related Work and Background 48
2.1 Optimised Sequential Decision Making 49
2.2 Contribution 50
3 Methodology 51
3.1 Optimal Postponing Policy 51
3.2 Policies Under Comparison 54
4 Performance Evaluation 58
4.1 Data Sets 58
4.2 Experimentation with Linear Regression Models 58
4.3 Experimentation with Support Vector Regression Models 61
4.4 Evaluation Summary 65
5 Conclusions 66
References 71
Intelligent Vehicular Networking Protocols 73
1 Introduction 74
2 Routing Protocols in VANET 76
2.1 Topology Based Routing Protocols 77
2.2 Position-Based Routing Protocols or Geographic Routing Protocols 84
2.3 Broadcast Routing 91
2.4 Geocast Routing Protocols 93
2.5 Cluster-Based Routing Protocols 95
3 Internet of Vehicles 97
3.1 Unicast Protocol 98
3.2 Multicast Protocol 98
3.3 Broadcast Protocol 99
4 Conclusion 99
References 99
Towards Ubiquitous Privacy Decision Support: Machine Prediction of Privacy Decisions in IoT 101
1 Introduction 101
2 Related Work 104
2.1 Prediction of Privacy Decision-Making 105
2.2 Privacy Segmentation 107
3 Dataset of Privacy Decisions 108
4 Privacy Decision Prediction 109
4.1 Machine Learning Model 110
4.2 Features 112
4.3 Training Strategy 116
4.4 Implications 120
5 Discussion and Future Work 123
5.1 Representability of Data 123
5.2 Reliability of Privacy Segmentation 123
5.3 Privacy Paradox 124
6 Conclusion 124
References 127
Energy-Efficient Design of Data Center Spaces in the Era of IoT Exploiting the Concept of Digital Twins 130
1 Introduction 131
2 Related Work 132
2.1 Energy Consumption in Data Centers 132
2.2 Degrees of Freedom in Energy Efficiency 135
2.3 Power Usage Effectiveness 137
2.4 Data Centers and Building Envelopes 139
3 Proposed Methodology 140
3.1 Modeling Procedure 140
3.2 Software Simulation Tools 142
3.3 Creating the 3D Geometry in SketchUp 143
3.4 Measurements 145
3.5 Setting Model Parameters in OpenStudio .OSM File 146
3.6 Hardware Laboratory’s Simulation Results 149
3.7 Data Center Simulation Results 150
4 Models’ Validation 151
4.1 Hardware Laboratory’s Validation Results 151
4.2 Data Center’s Validation Results 152
4.3 PUE Calculations for Various Structural Interventions 152
5 Conclusions 153
References 155
In-Network Machine Learning Predictive Analytics: A Swarm Intelligence Approach 157
1 Introduction 158
1.1 Motivation 158
1.2 Aim 159
1.3 Outline 159
2 Background 160
2.1 Particle Swarm Optimisation Algorithm 160
2.2 Particle Representation 161
2.3 Parameters 162
2.4 The PSO Algorithm 163
2.5 Regression 164
2.6 Prediction Error Metrics 166
2.7 Network Modelling 166
2.8 Mica2 Wireless Sensor Platform 168
3 Analysis 169
3.1 Problem Definition and Analysis 170
3.2 Baseline Methodology 171
3.3 Limitations 172
3.4 Proposed Methodology 173
3.5 Network Model Impact 175
4 Implementation 175
4.1 Programming Language 175
4.2 Data 176
4.3 Particle Swarm Optimisation 178
4.4 Convergence 181
4.5 Network Models 182
5 Performance and Comparative Assessment 184
5.1 Prerequisites 184
5.2 Random Network Assessment 185
5.3 Small World Network Assessment 189
6 Conclusions and Future Work 191
6.1 Assessment Results 191
6.2 Future Work 194
6.3 Summary 194
References 195
Machine Learning Techniques for Wireless-Powered Ambient Backscatter Communications: Enabling Intelligent IoT Networks in 6G Era 198
1 Introduction to Artificial Intelligence (AI) 199
2 Machine Learning Paradigm and Techniques 201
2.1 Supervised Learning 202
2.2 Unsupervised Learning 203
2.3 Reinforcement Learning 203
3 Robustness of Deep (Machine) Learning Approaches 204
4 Hardware Requirements for Implementing Machine Learning Techniques 206
4.1 Computational Cost: CPU Vs GPU 206
4.2 Existing Hardware Solutions 208
5 Ambient Backscatter Communications and Machine Learning 209
5.1 Basics of Backscatter Communications 210
5.2 Applications of Machine Learning in Ambient Backscatter Communications 210
5.3 Selection of Reflection Coefficient 212
6 System Model 213
7 Power Control in Ambient Backscatter Communications 215
8 Simulation Results and Discussion 217
9 Conclusion and Future Research Directions 219
References 220
Processing Systems for Deep Learning Inference on Edge Devices 223
1 Introduction 223
2 Deep Learning 225
3 Complexity Reduction of Deep Learning Models 228
4 Deep Learning Computation 230
5 Computing Platforms for Edge Devices 233
5.1 Reconfigurable Platforms for Edge Computing 234
5.2 ASICs for Edge Computing 241
6 Inference at the Edge: Present and Future 245
7 Conclusions 247
References 248
Power Domain Based Multiple Access for IoT Deployment: Two-Way Transmission Mode and Performance Analysis 251
1 Introduction 252
2 System Model 253
3 Performance Evaluation: Outage Performance Analysis 255
3.1 The Outage Probability of the First User Pair 255
4 Numerical Results 258
5 Conclusion 260
References 267
Big Data Thinning: Knowledge Discovery from Relevant Data 269
1 Introduction 269
1.1 Motivation 270
1.2 Aims and Hypotheses 271
1.3 Performance Assessment 271
1.4 Outline 271
2 Background and Relevant Research 272
2.1 Competitive Learning 272
2.2 Learning Automata with Two Actions 276
2.3 The Improved Initialisation Algorithm: K-Means++ 277
3 Hypothesis 1 278
3.1 Design 279
3.2 Implementation 280
3.3 Evaluation 285
4 Hypothesis 2 290
4.1 Design 290
5 Implementation 292
5.1 Generating Queries/Building the Query Space 292
5.2 Query Quantisation Using RPCL 294
5.3 Evaluation 298
6 Conclusions and Future Work 304
6.1 Generalisation of Findings 304
6.2 Future Work 305
References 306
Optimizing Blockchain Networks with Artificial Intelligence: Towards Efficient and Reliable IoT Applications 308
1 Blockchain 309
1.1 Blocks 310
1.2 Transaction 311
1.3 Signing a Transaction 311
1.4 Mining 313
2 Applications of Blockchain Technology 314
2.1 Local Energy Trading 314
2.2 Internet of Things 315
2.3 Next Generation Payment Solutions 316
3 Fundamentals of Artificial Intelligence 316
3.1 Artificial Intelligence in Modern Era 318
3.2 Impact of Artificial Intelligence on Industries 320
4 Data and Infrastructure Desideratum of Artificial Intelligence 321
4.1 IT Infrastructures 321
4.2 Algorithms and Methods 321
4.3 Training Data 322
5 Blockchain Network Architecture 322
6 System Model 324
7 Optimization Using Deep Neural Networks 325
7.1 Problem Formulation 325
7.2 Deep Learning Network Setup 326
8 Numerical Results 326
9 Conclusion 328
References 328
Industrial and Artificial Internet of Things with Augmented Reality 331
1 Introduction 331
1.1 Artificial Intelligence 332
1.2 Augmented Reality 333
2 Industrial Internet of Things 334
2.1 Cyber-Physical Systems 336
2.2 Industrial Augmented Reality 337
2.3 IoT Communication Systems 340
3 Artificial Internet of Things 341
3.1 Bringing Intelligence to the Data 342
3.2 Edge and Fog Computing 344
4 Applications 346
4.1 Manufacturing Sector 346
4.2 Aerospace Sector 348
4.3 Logistics Sector 348
4.4 Water Sector 349
5 Future Directions 351
6 Conclusion 352
References 353
IoT Detection Techniques for Modeling Post-Fire Landscape Alteration Using Multitemporal Spectral Indices 355
1 Introduction 356
1.1 Remote Sensing: Delivering on the IoT 356
1.2 Study Area 356
2 Materials and Methods 357
2.1 Area of Interest 357
2.2 Remote Sensors on Landsat 8 OLI/TIRS 358
2.3 Dataset 359
2.4 Classification 359
2.5 Variation Detection Processing 360
3 Results Verification 367
3.1 Results 368
3.2 Significance of Results 370
4 Conclusion 373
References 374
Internet of Things and Artificial Intelligence—A Wining Partnership? 376
1 Introduction 377
2 Technology Background—Five Decades that Changed the Electrical Engineering Paradigm 379
3 Human Background 382
4 Challenges of IoT-AI Partnership 384
5 Towards an Intelligent Strategy for AI-Driven IoT. Areas of Urgent Research 389
5.1 Engineering Education 389
5.2 Design for Accountability (DfA) 391
6 Conclusions 395
References 396
AI Architectures for Very Smart Sensors 398
1 Introduction 398
2 Key Characteristics of Artificial Neural Networks 402
2.1 Hierarchy of Feature Detectors 402
2.2 Fully Connected Layers in Neural Networks 402
2.3 Convolutional Layers in Neural Networks 403
2.4 Vanishing or Exploding Gradient in Connection with Stochastic Gradient Descent 405
2.5 Shortcut Connections 406
2.6 Transfer Learning 407
3 Architectures of Modern Neural Networks 408
3.1 LeNet-5 408
3.2 AlexNet 409
3.3 VGGNet 410
3.4 GoogLeNet 411
3.5 ResNet 414
3.6 U-Net 416
3.7 DenseNet 417
3.8 PolyNet 418
3.9 ResNeXt 420
3.10 DualPathNet 420
4 Classification Accuracy of Modern Neural Networks 421
5 Computational Complexity Reducing Techniques of Modern Neural Networks 426
5.1 Reducing the Depth and Width of Neural Networks 426
5.2 Reducing the Resolution of the Input 426
5.3 More Efficient Building Blocks of Neural Networks 427
5.4 Separable Convolution 427
5.5 Depth-Wise Separable Convolution 427
5.6 Pruning 428
5.7 Compression of Parameters 428
5.8 Reduced Floating-Point Precision 429
5.9 Quantification of Weights into Integer or Binary Values 429
5.10 Quantification of Both Weights and Activations into Integer or Binary Values 430
5.11 More Efficient Computational Hardware 431
5.12 Custom Hardware Design 431
6 Neural Networks for Mobile Devices and IoT 432
6.1 SqueezeNet 432
6.2 MobileNet 433
6.3 ShuffleNet 434
6.4 LQ-Nets 435
6.5 GroupNet 436
7 Classification Accuracy of Neural Networks for Mobile and IoT 437
8 Conclusion 441
References 442