Daniel Minoli is Principal Consultant for DVI Communications, New York, USA, and a longtime Expert Witness and Testifying Expert in networking, wireless, video, IoT, and VoIP. In addition to working as Director of Engineering for gamut of premiere high-tech firms, he has acted as Adjunct Instructor at New York University and Stevens Institute of Technology, USA for twenty years. He has published extensively on networks, IP/IPv6, video, wireless communications, and related subjects. Benedict Occhiogrosso is Co-Founder of DVI Communications, New York, USA, with extensive experience as a technology engineer, manager and executive. He is a subject matter expert in several disciplines now enhanced by artificial intelligence including telecommunications networking, speech recognition, image processing and building management systems. He has also served as a testifying expert witness and advisor on patent portfolios.

• About the Authors xiPreface xiii1 Overview 11.1 Introduction and Basic Concepts 11.1.1 Machine Learning 51.1.2 Deep Learning 61.1.3 Activation Functions 131.1.4 Multi-layer Perceptrons 171.1.5 Recurrent Neural Networks 211.1.6 Convolutional Neural Networks 211.1.7 Comparison 261.2 Learning Methods 261.3 Areas of Applicability 391.4 Scope of this Text 41A. Basic Glossary of Key AI Terms and Concepts 44References 572 Current and Evolving Applications to Natural Language Processing 652.1 Scope 652.2 Introduction 662.3 Overview of Natural Language Processing and Speech Processing 722.3.1 Feed-forward NNs 742.3.2 Recurrent Neural Networks 742.3.3 Long Short-Term Memory 752.3.4 Attention 772.3.5 Transformer 782.4 Natural Language Processing/Natural Language Understanding Basics 812.4.1 Pre-training 822.4.2 Natural Language Processing/Natural Language Generation Architectures 852.4.3 Encoder-Decoder Methods 882.4.4 Application of Transformer 892.4.5 Other Approaches 902.5 Natural Language Generation Basics 912.6 Chatbots 952.7 Generative AI 101A. Basic Glossary of Key AI Terms and Concepts Related to Natural Language Processing 103References 1093 Current and Evolving Applications to Speech Processing 1173.1 Scope 1173.2 Overview 1193.2.1 Traditional Approaches 1193.2.2 DNN-based Feature Extraction 1233.3 Noise Cancellation 1263.3.1 Approaches 1283.3.1.1 Delay-and-Sum Beamforming (DSB) 1293.3.1.2 Minimum Variance Distortionless Response (MVDR) Beamformer 1303.3.1.3 Non- adaptive Beamformer 1313.3.1.4 Multichannel Linear Prediction (MCLP) 1323.3.1.5 ML-based Approaches 1323.3.1.6 Neural Network Beamforming 1353.3.2 Specific Example of a System Supporting Noise Cancellation 1383.4 Training 1413.5 Applications to Voice Interfaces Used to Control Home Devices and Digital Assistant Applications 1423.6 Attention-based Models 1463.7 Sentiment Extraction 1483.8 End-to-End Learning 1483.9 Speech Synthesis 1503.10 Zero-shot TTS 1523.11 VALL- E: Unseen Speaker as an Acoustic Prompt 152A. Basic Glossary of Key AI Terms and Concepts 156References 1664 Current and Evolving Applications to Video and Imaging 1734.1 Overview and Background 1734.2 Convolution Process 1764.3 CNNs 1814.3.1 Nomenclature 1814.3.2 Basic Formulation of the CNN Layers and Operation 1814.3.2.1 Layers 1814.3.2.2 Operations 1884.3.3 Fully Convolutional Networks (FCN) 1904.3.4 Convolutional Autoencoders 1904.3.5 R-CNNs, Fast R-CNN, Faster R-CNN 1934.4 Imaging Applications 1954.4.1 Basic Image Management 1954.4.2 Image Segmentation and Image Classification 1994.4.3 Illustrative Examples of a Classification DNN/CNN 2024.4.4 Well-Known Image Classification Networks 2044.5 Specific Application Examples 2134.5.1 Semantic Segmentation and Semantic Edge Detection 2134.5.2 CNN Filtering Process for Video Coding 2154.5.3 Virtual Clothing 2164.5.4 Example of Unmanned Underwater Vehicles/Unmanned Aerial Vehicles 2184.5.5 Object Detection Applications 2184.5.6 Classifying Video Data 2224.5.7 Example of Training 2244.5.8 Example: Image Reconstruction is Used to Remove Artifacts 2254.5.9 Example: Video Transcoding/Resolution-enhancement 2284.5.10 Facial Expression Recognition 2284.5.11 Transformer Architecture for Image Processing 2304.5.12 Example: A GAN Approach/Synthetic Photo 2304.5.13 Situational Awareness 2314.6 Other Models: Diffusion and Consistency Models 236A. Basic Glossary of Key AI Terms and Concepts 238B. Examples of Convolutions 246References 2505 Current and Evolving Applications to IoT and Applications to Smart Buildings and Energy Management 2575.1 Introduction 2575.1.1 IoT Applications 2575.1.2 Smart Cities 2585.2 Smart Building ML Applications 2755.2.1 Basic Building Elements 2755.2.2 Particle Swarm Optimization 2765.2.3 Specific ML Example – Qin Model 2795.2.3.1 EnergyPlus™ 2815.2.3.2 Modeling and Simulation 2825.2.3.3 Energy Audit Stage 2865.2.3.4 Optimization Stage 2875.2.3.5 Model Construction 2895.2.3.6 EnergyPlus Models 2895.2.3.7 Real- Time Control Parameters 2905.2.3.8 Neural Networks in the Qin Model (DNN, RNN, CNN) 2905.2.3.9 Finding Inefficiency Measures 2945.2.3.10 Particle Swarm Optimizer 2945.2.3.11 Integration of Particle Swarm Optimization with Neural Networks 2965.2.3.12 Deep Reinforcement Learning 2985.2.3.13 Deployments 2985.3 Example of a Commercial Product – BrainBox AI 3015.3.1 Overview 3015.3.2 LSTM Application – Technical Background 3025.3.3 BrainBox AI Commercial Energy Optimization System 305A. Basic Glossary of Key IoT (Smart Building) Terms and Concepts 314References 3396 Current and Evolving Applications to Network Cybersecurity 3476.1 Overview 3476.2 General Security Requirements 3496.3 Corporate Resources/Intranet Security Requirements 3536.3.1 Network and End System Security Testing 3586.3.2 Application Security Testing 3606.3.3 Compliance Testing 3626.4 IoT Security (IoTSec) 3636.5 Blockchains 3656.6 Zero Trust Environments 3696.7 Areas of ML Applicability 3706.7.1 Example of Cyberintrusion Detector 3736.7.2 Example of Hidden Markov Model (HMM) for Intrusion Detection 3746.7.3 Anomaly Detection Example 3786.7.4 Phishing Detection Emails Using Feature Extraction 3836.7.5 Example of Classifier Engine to Identify Phishing Websites 3866.7.6 Example of System for Data Protection 3886.7.7 Example of an Integrated Cybersecurity Threat Management 3906.7.8 Example of a Vulnerability Lifecycle Management System 392A. Basic Glossary of Key Security Terms and Concepts 396References 4007 Current and Evolving Applications to Network Management 4077.1 Overview 4077.2 Examples of Neural Network- Assisted Network Management 4087.2.1 Example of NN-Based Network Management System (Case of FM) 4137.2.2 Example of a Model for Predictions Related to the Operation of a Telecommunication Network (Case of FM) 4167.2.3 Prioritizing Network Monitoring Alerts (Case of FM and PM) 4197.2.4 System for Recognizing and Addressing Network Alarms (Case of FM) 4247.2.5 Load Control of an Enterprise Network (Case of PM) 4287.2.6 Data Reduction to Accelerate Machine Learning for Networking (Case of FM and PM) 4317.2.7 Compressing Network Data (Case of PM) 4357.2.8 ML Predictor for a Remote Network Management Platform (Case of FM, PM, CM, AM) 4377.2.9 Cable Television (CATV) Performance Management System (Case of PM) 441A. Short Glossary of Network Management Concepts 446References 447Super Glossary 449Index 467

Daniel Minoli is Principal Consultant for DVI Communications, New York, USA, and a longtime Expert Witness and Testifying Expert in networking, wireless, video, IoT, and VoIP. In addition to working as Director of Engineering for gamut of premiere high-tech firms, he has acted as Adjunct Instructor at New York University and Stevens Institute of Technology, USA for twenty years. He has published extensively on networks, IP/IPv6, video, wireless communications, and related subjects. Benedict Occhiogrosso is Co-Founder of DVI Communications, New York, USA, with extensive experience as a technology engineer, manager and executive. He is a subject matter expert in several disciplines now enhanced by artificial intelligence including telecommunications networking, speech recognition, image processing and building management systems. He has also served as a testifying expert witness and advisor on patent portfolios.

• About the Authors xiPreface xiii1 Overview 11.1 Introduction and Basic Concepts 11.1.1 Machine Learning 51.1.2 Deep Learning 61.1.3 Activation Functions 131.1.4 Multi-layer Perceptrons 171.1.5 Recurrent Neural Networks 211.1.6 Convolutional Neural Networks 211.1.7 Comparison 261.2 Learning Methods 261.3 Areas of Applicability 391.4 Scope of this Text 41A. Basic Glossary of Key AI Terms and Concepts 44References 572 Current and Evolving Applications to Natural Language Processing 652.1 Scope 652.2 Introduction 662.3 Overview of Natural Language Processing and Speech Processing 722.3.1 Feed-forward NNs 742.3.2 Recurrent Neural Networks 742.3.3 Long Short-Term Memory 752.3.4 Attention 772.3.5 Transformer 782.4 Natural Language Processing/Natural Language Understanding Basics 812.4.1 Pre-training 822.4.2 Natural Language Processing/Natural Language Generation Architectures 852.4.3 Encoder-Decoder Methods 882.4.4 Application of Transformer 892.4.5 Other Approaches 902.5 Natural Language Generation Basics 912.6 Chatbots 952.7 Generative AI 101A. Basic Glossary of Key AI Terms and Concepts Related to Natural Language Processing 103References 1093 Current and Evolving Applications to Speech Processing 1173.1 Scope 1173.2 Overview 1193.2.1 Traditional Approaches 1193.2.2 DNN-based Feature Extraction 1233.3 Noise Cancellation 1263.3.1 Approaches 1283.3.1.1 Delay-and-Sum Beamforming (DSB) 1293.3.1.2 Minimum Variance Distortionless Response (MVDR) Beamformer 1303.3.1.3 Non- adaptive Beamformer 1313.3.1.4 Multichannel Linear Prediction (MCLP) 1323.3.1.5 ML-based Approaches 1323.3.1.6 Neural Network Beamforming 1353.3.2 Specific Example of a System Supporting Noise Cancellation 1383.4 Training 1413.5 Applications to Voice Interfaces Used to Control Home Devices and Digital Assistant Applications 1423.6 Attention-based Models 1463.7 Sentiment Extraction 1483.8 End-to-End Learning 1483.9 Speech Synthesis 1503.10 Zero-shot TTS 1523.11 VALL- E: Unseen Speaker as an Acoustic Prompt 152A. Basic Glossary of Key AI Terms and Concepts 156References 1664 Current and Evolving Applications to Video and Imaging 1734.1 Overview and Background 1734.2 Convolution Process 1764.3 CNNs 1814.3.1 Nomenclature 1814.3.2 Basic Formulation of the CNN Layers and Operation 1814.3.2.1 Layers 1814.3.2.2 Operations 1884.3.3 Fully Convolutional Networks (FCN) 1904.3.4 Convolutional Autoencoders 1904.3.5 R-CNNs, Fast R-CNN, Faster R-CNN 1934.4 Imaging Applications 1954.4.1 Basic Image Management 1954.4.2 Image Segmentation and Image Classification 1994.4.3 Illustrative Examples of a Classification DNN/CNN 2024.4.4 Well-Known Image Classification Networks 2044.5 Specific Application Examples 2134.5.1 Semantic Segmentation and Semantic Edge Detection 2134.5.2 CNN Filtering Process for Video Coding 2154.5.3 Virtual Clothing 2164.5.4 Example of Unmanned Underwater Vehicles/Unmanned Aerial Vehicles 2184.5.5 Object Detection Applications 2184.5.6 Classifying Video Data 2224.5.7 Example of Training 2244.5.8 Example: Image Reconstruction is Used to Remove Artifacts 2254.5.9 Example: Video Transcoding/Resolution-enhancement 2284.5.10 Facial Expression Recognition 2284.5.11 Transformer Architecture for Image Processing 2304.5.12 Example: A GAN Approach/Synthetic Photo 2304.5.13 Situational Awareness 2314.6 Other Models: Diffusion and Consistency Models 236A. Basic Glossary of Key AI Terms and Concepts 238B. Examples of Convolutions 246References 2505 Current and Evolving Applications to IoT and Applications to Smart Buildings and Energy Management 2575.1 Introduction 2575.1.1 IoT Applications 2575.1.2 Smart Cities 2585.2 Smart Building ML Applications 2755.2.1 Basic Building Elements 2755.2.2 Particle Swarm Optimization 2765.2.3 Specific ML Example – Qin Model 2795.2.3.1 EnergyPlus™ 2815.2.3.2 Modeling and Simulation 2825.2.3.3 Energy Audit Stage 2865.2.3.4 Optimization Stage 2875.2.3.5 Model Construction 2895.2.3.6 EnergyPlus Models 2895.2.3.7 Real- Time Control Parameters 2905.2.3.8 Neural Networks in the Qin Model (DNN, RNN, CNN) 2905.2.3.9 Finding Inefficiency Measures 2945.2.3.10 Particle Swarm Optimizer 2945.2.3.11 Integration of Particle Swarm Optimization with Neural Networks 2965.2.3.12 Deep Reinforcement Learning 2985.2.3.13 Deployments 2985.3 Example of a Commercial Product – BrainBox AI 3015.3.1 Overview 3015.3.2 LSTM Application – Technical Background 3025.3.3 BrainBox AI Commercial Energy Optimization System 305A. Basic Glossary of Key IoT (Smart Building) Terms and Concepts 314References 3396 Current and Evolving Applications to Network Cybersecurity 3476.1 Overview 3476.2 General Security Requirements 3496.3 Corporate Resources/Intranet Security Requirements 3536.3.1 Network and End System Security Testing 3586.3.2 Application Security Testing 3606.3.3 Compliance Testing 3626.4 IoT Security (IoTSec) 3636.5 Blockchains 3656.6 Zero Trust Environments 3696.7 Areas of ML Applicability 3706.7.1 Example of Cyberintrusion Detector 3736.7.2 Example of Hidden Markov Model (HMM) for Intrusion Detection 3746.7.3 Anomaly Detection Example 3786.7.4 Phishing Detection Emails Using Feature Extraction 3836.7.5 Example of Classifier Engine to Identify Phishing Websites 3866.7.6 Example of System for Data Protection 3886.7.7 Example of an Integrated Cybersecurity Threat Management 3906.7.8 Example of a Vulnerability Lifecycle Management System 392A. Basic Glossary of Key Security Terms and Concepts 396References 4007 Current and Evolving Applications to Network Management 4077.1 Overview 4077.2 Examples of Neural Network- Assisted Network Management 4087.2.1 Example of NN-Based Network Management System (Case of FM) 4137.2.2 Example of a Model for Predictions Related to the Operation of a Telecommunication Network (Case of FM) 4167.2.3 Prioritizing Network Monitoring Alerts (Case of FM and PM) 4197.2.4 System for Recognizing and Addressing Network Alarms (Case of FM) 4247.2.5 Load Control of an Enterprise Network (Case of PM) 4287.2.6 Data Reduction to Accelerate Machine Learning for Networking (Case of FM and PM) 4317.2.7 Compressing Network Data (Case of PM) 4357.2.8 ML Predictor for a Remote Network Management Platform (Case of FM, PM, CM, AM) 4377.2.9 Cable Television (CATV) Performance Management System (Case of PM) 441A. Short Glossary of Network Management Concepts 446References 447Super Glossary 449Index 467

Daniel Minoli is Principal Consultant for DVI Communications, New York, USA, and a longtime Expert Witness and Testifying Expert in networking, wireless, video, IoT, and VoIP. In addition to working as Director of Engineering for gamut of premiere high-tech firms, he has acted as Adjunct Instructor at New York University and Stevens Institute of Technology, USA for twenty years. He has published extensively on networks, IP/IPv6, video, wireless communications, and related subjects. Benedict Occhiogrosso is Co-Founder of DVI Communications, New York, USA, with extensive experience as a technology engineer, manager and executive. He is a subject matter expert in several disciplines now enhanced by artificial intelligence including telecommunications networking, speech recognition, image processing and building management systems. He has also served as a testifying expert witness and advisor on patent portfolios.

• About the Authors xiPreface xiii1 Overview 11.1 Introduction and Basic Concepts 11.1.1 Machine Learning 51.1.2 Deep Learning 61.1.3 Activation Functions 131.1.4 Multi-layer Perceptrons 171.1.5 Recurrent Neural Networks 211.1.6 Convolutional Neural Networks 211.1.7 Comparison 261.2 Learning Methods 261.3 Areas of Applicability 391.4 Scope of this Text 41A. Basic Glossary of Key AI Terms and Concepts 44References 572 Current and Evolving Applications to Natural Language Processing 652.1 Scope 652.2 Introduction 662.3 Overview of Natural Language Processing and Speech Processing 722.3.1 Feed-forward NNs 742.3.2 Recurrent Neural Networks 742.3.3 Long Short-Term Memory 752.3.4 Attention 772.3.5 Transformer 782.4 Natural Language Processing/Natural Language Understanding Basics 812.4.1 Pre-training 822.4.2 Natural Language Processing/Natural Language Generation Architectures 852.4.3 Encoder-Decoder Methods 882.4.4 Application of Transformer 892.4.5 Other Approaches 902.5 Natural Language Generation Basics 912.6 Chatbots 952.7 Generative AI 101A. Basic Glossary of Key AI Terms and Concepts Related to Natural Language Processing 103References 1093 Current and Evolving Applications to Speech Processing 1173.1 Scope 1173.2 Overview 1193.2.1 Traditional Approaches 1193.2.2 DNN-based Feature Extraction 1233.3 Noise Cancellation 1263.3.1 Approaches 1283.3.1.1 Delay-and-Sum Beamforming (DSB) 1293.3.1.2 Minimum Variance Distortionless Response (MVDR) Beamformer 1303.3.1.3 Non- adaptive Beamformer 1313.3.1.4 Multichannel Linear Prediction (MCLP) 1323.3.1.5 ML-based Approaches 1323.3.1.6 Neural Network Beamforming 1353.3.2 Specific Example of a System Supporting Noise Cancellation 1383.4 Training 1413.5 Applications to Voice Interfaces Used to Control Home Devices and Digital Assistant Applications 1423.6 Attention-based Models 1463.7 Sentiment Extraction 1483.8 End-to-End Learning 1483.9 Speech Synthesis 1503.10 Zero-shot TTS 1523.11 VALL- E: Unseen Speaker as an Acoustic Prompt 152A. Basic Glossary of Key AI Terms and Concepts 156References 1664 Current and Evolving Applications to Video and Imaging 1734.1 Overview and Background 1734.2 Convolution Process 1764.3 CNNs 1814.3.1 Nomenclature 1814.3.2 Basic Formulation of the CNN Layers and Operation 1814.3.2.1 Layers 1814.3.2.2 Operations 1884.3.3 Fully Convolutional Networks (FCN) 1904.3.4 Convolutional Autoencoders 1904.3.5 R-CNNs, Fast R-CNN, Faster R-CNN 1934.4 Imaging Applications 1954.4.1 Basic Image Management 1954.4.2 Image Segmentation and Image Classification 1994.4.3 Illustrative Examples of a Classification DNN/CNN 2024.4.4 Well-Known Image Classification Networks 2044.5 Specific Application Examples 2134.5.1 Semantic Segmentation and Semantic Edge Detection 2134.5.2 CNN Filtering Process for Video Coding 2154.5.3 Virtual Clothing 2164.5.4 Example of Unmanned Underwater Vehicles/Unmanned Aerial Vehicles 2184.5.5 Object Detection Applications 2184.5.6 Classifying Video Data 2224.5.7 Example of Training 2244.5.8 Example: Image Reconstruction is Used to Remove Artifacts 2254.5.9 Example: Video Transcoding/Resolution-enhancement 2284.5.10 Facial Expression Recognition 2284.5.11 Transformer Architecture for Image Processing 2304.5.12 Example: A GAN Approach/Synthetic Photo 2304.5.13 Situational Awareness 2314.6 Other Models: Diffusion and Consistency Models 236A. Basic Glossary of Key AI Terms and Concepts 238B. Examples of Convolutions 246References 2505 Current and Evolving Applications to IoT and Applications to Smart Buildings and Energy Management 2575.1 Introduction 2575.1.1 IoT Applications 2575.1.2 Smart Cities 2585.2 Smart Building ML Applications 2755.2.1 Basic Building Elements 2755.2.2 Particle Swarm Optimization 2765.2.3 Specific ML Example – Qin Model 2795.2.3.1 EnergyPlus™ 2815.2.3.2 Modeling and Simulation 2825.2.3.3 Energy Audit Stage 2865.2.3.4 Optimization Stage 2875.2.3.5 Model Construction 2895.2.3.6 EnergyPlus Models 2895.2.3.7 Real- Time Control Parameters 2905.2.3.8 Neural Networks in the Qin Model (DNN, RNN, CNN) 2905.2.3.9 Finding Inefficiency Measures 2945.2.3.10 Particle Swarm Optimizer 2945.2.3.11 Integration of Particle Swarm Optimization with Neural Networks 2965.2.3.12 Deep Reinforcement Learning 2985.2.3.13 Deployments 2985.3 Example of a Commercial Product – BrainBox AI 3015.3.1 Overview 3015.3.2 LSTM Application – Technical Background 3025.3.3 BrainBox AI Commercial Energy Optimization System 305A. Basic Glossary of Key IoT (Smart Building) Terms and Concepts 314References 3396 Current and Evolving Applications to Network Cybersecurity 3476.1 Overview 3476.2 General Security Requirements 3496.3 Corporate Resources/Intranet Security Requirements 3536.3.1 Network and End System Security Testing 3586.3.2 Application Security Testing 3606.3.3 Compliance Testing 3626.4 IoT Security (IoTSec) 3636.5 Blockchains 3656.6 Zero Trust Environments 3696.7 Areas of ML Applicability 3706.7.1 Example of Cyberintrusion Detector 3736.7.2 Example of Hidden Markov Model (HMM) for Intrusion Detection 3746.7.3 Anomaly Detection Example 3786.7.4 Phishing Detection Emails Using Feature Extraction 3836.7.5 Example of Classifier Engine to Identify Phishing Websites 3866.7.6 Example of System for Data Protection 3886.7.7 Example of an Integrated Cybersecurity Threat Management 3906.7.8 Example of a Vulnerability Lifecycle Management System 392A. Basic Glossary of Key Security Terms and Concepts 396References 4007 Current and Evolving Applications to Network Management 4077.1 Overview 4077.2 Examples of Neural Network- Assisted Network Management 4087.2.1 Example of NN-Based Network Management System (Case of FM) 4137.2.2 Example of a Model for Predictions Related to the Operation of a Telecommunication Network (Case of FM) 4167.2.3 Prioritizing Network Monitoring Alerts (Case of FM and PM) 4197.2.4 System for Recognizing and Addressing Network Alarms (Case of FM) 4247.2.5 Load Control of an Enterprise Network (Case of PM) 4287.2.6 Data Reduction to Accelerate Machine Learning for Networking (Case of FM and PM) 4317.2.7 Compressing Network Data (Case of PM) 4357.2.8 ML Predictor for a Remote Network Management Platform (Case of FM, PM, CM, AM) 4377.2.9 Cable Television (CATV) Performance Management System (Case of PM) 441A. Short Glossary of Network Management Concepts 446References 447Super Glossary 449Index 467