Democratization of Artificial Intelligence for the Future of Humanity
- Length: 388 pages
- Edition: 1
- Language: English
- Publisher: CRC Press
- Publication Date: 2021-01-18
- ISBN-10: 0367524120
- ISBN-13: 9780367524128
- Sales Rank: #880980 (See Top 100 Books)
Artificial intelligence (AI) stands out as a transformational technology of the digital age. Its practical applications are growing very rapidly. One of the chief reasons AI applications are attaining prominence, is in its design to learn continuously, from real-world use and experience, and its capability to improve its performance. It is no wonder that the applications of AI span from complex high-technology equipment manufacturing to personalized exclusive recommendations to end-users. Many deployments of AI software, given its continuous learning need, require computation platforms that are resource intense, and have sustained connectivity and perpetual power through central electrical grid.
In order to harvest the benefits of AI revolution to all of humanity, traditional AI software development paradigms must be upgraded to function effectively in environments that have resource constraints, small form factor computational devices with limited power, devices with intermittent or no connectivity and/or powered by non-perpetual source or battery power.
The aim this book is to prepare current and future software engineering teams with the skills and tools to fully utilize AI capabilities in resource-constrained devices. The book introduces essential AI concepts from the perspectives of full-scale software development with emphasis on creating niche Blue Ocean small form factored computational environment products.
Cover Title Page Copyright Page Dedication Preface Acknowledgements Contents Section I—Introduction to Artificial Intelligence & Frameworks 1. Introduction What is AI? Machine Learning Types of Analytics Descriptive Analytics Predictive Analytics Human vs. BOT Web Traffic Prediction ML Use Case Prescriptive Analytics Construction Management and Prescriptive Analytics Use Case ML and Types of data Structured data Unstructured data Semi-structured data Machine Learning and Large-Scale Analytics Big Data Example of Large-scale Analytical Systems Types of Learning Eager Learner vs. Lazy Learner Supervised Learning Unsupervised Learning Reinforcement Learning Deep Learning What Is a Heuristic? Choosing the Right Estimator Mapping AI Technique to Classical ML AI Epochs: Waves of Compute First Wave of Computing Second Wave of Computing Third Wave of Computing The Cray 2 Computer System vs. iPhone XS Comparison of CRAY-2 vs. iPhone XS Fourth Wave of Computing Fifth Wave of Computing AI Hype Cycle—Current and Emerging Technologies Hype Cycle for AI, 2017 Hype Cycle for AI, 2018 Hype Cycle for AI, 2019 Digital Strategy AI End-To-End (E2E) Process—Turning Data into Actionable Insights Data Science Lifecycle Data Sources Prepare & Transform Exploratory Data Analysis (EDA) Model Visualization Microsoft Azure—AI E2E Platform AI Development Operations (DevOps) Loop for Data Science Data Preparation Experiment Build Model Train & Test Model Deployment Edge Devices ML Constrained Modeling Constrained IoT Edge Devices Infrastructure Constraints Operating Environment Device Characteristics AI Performance and Computational Notations Sample complexity AI Algorithm and Computational Complexity Analysis Time Complexity Space Complexity Algorithm Performance Metrics—The Asymptotic Notions Big-O Notation (O-Notation) Omega Notation (Ω Notation) Theta Notation (Θ Notation) Space and Time constraint benchmarks AI for Greater Good—Solving Humanity and Societal Challenges References 2. Standard Processes and Frameworks Digital Transformation Digital Transformation at Salesforce Definition Digital Feedback Loop Insights Value Chain Data Analytics IT People Process Strategy and Vision Operating Model The CRISP-DM Process Business Understanding Data Understanding Data Preparation Phase Data Modelling Best to Divide data into Training and Validation Data Holdout Data Stratification Deployment Building Blocks of AI—Major Components of AI AI Reference Architectures Knowledge Discovery in Databases (KDD) Data Mining Reference Architecture Streaming Processing Reference Architecture Stream Data Examples Sensor data Image data Internet and Web Traffic data Streaming Processing Rules BLAST—Stream Model AI Data Pipeline Edge Processing End-to-End Platform Data Shifting to the Edge Harsh Conditions Remote Locations Quality of Service (QoS)/Low Latency No-Touch/Self-Healing Global Data Experience Nomenclature of Embedded and Edge Devices Microcontrollers [28] Connectivity Operating System Power Optimization References Section II—Data Sources and Engineering Tools 3. Data—Call for Democratization Call for Action Creating Sustainable Food Future by 2050 Nitrogen and Phosphorus Pollution Data Access Tool (NPDAT) Investments Designing AI that Uses Less Energy The Last Mile—Constrained Compute Devices & "AI Chasm" Classes of Constrained Devices Class 0 devices Class 1 devices Class 2 devices Edge Device Architecture AI Model Custom Built Hardware ML Models ML Models based Packaged Frameworks—TensorFlow for C Linker Dependencies Connectivity Bluetooth Low Energy (BLE) Hardware—Storage Files in C EPROM Data Storage EPROM Read and Writes References 4. Machine Learning Frameworks and Device Engineering Machine Learning Device Deployments Extremely Resource-Constrained (xRC) Systems Deep Learning Device Deployment Arduino Nano 33 BLE Sense SparkFun Edge Adafruit EdgeBadge—TensorFlow Lite for Microcontrollers ESP32-DevKitC Extremely Resource-Constrained (xRC) Systems xRC Modeling: Model Accuracy-Connectivity-Hardware (MCH) Framework The Trade-off Modeling Hardware Economy—Model Accuracy—Connectivity Trade-off Hardware Economy Model (AI/ML) Accuracy Connectivity Hardware Economy—Model Accuracy trade-off Modeling no connectivity Modeling Low Power Bluetooth Connectivity The BLE Device Connect Offload data Modeling Wi-Fi Connectivity Connectivity—Model Accuracy trade-off Modeling Memory Memory Management Simple Memory Application IDE Output Circular Buffer Design Circular Buffers Circular Buffer C Code Modeling Power Storage Modeling AI Democratization—"Crossing the Chasm" Infrastructure Issues Connectivity AI Model Design Considerations: Connectivity Electricity Operating Environment AI Model Design Considerations: Operating Environmental Factors Variations in Targeted Platform Perturbations Thermal Characteristics Data Device Characteristics: Real-time, at the edge, and a Reasonable Price Point References 5. Device Software and Hardware Engineering Tools Software Engineering Tools Machine Learning Tools Anaconda Jupyter Notebook Spyder Android Studio Google Colaboratory Microsoft Azure Machine Learning Azure Databricks TensorFlow Install TensorFlow for C Supported Platforms TensorFlow Lite for Microcontrollers Hardware and Engineering Tools Eclipse IDE for C/C++ Developers Microsoft Visual Studio 2019 Cortex M3 Processor GNU Arm Embedded Toolchain Pre-built GNU Toolchain for Arm Cortex-M and Cortex-R Processors GNU C/C++ Compiler Doxygen Hyperload Libraries C Language – C11 GNU GCC Compiler The Compiler Options Microsoft Visual Studio—C++ Compiler MSVC Compiler Options Optimization References Section III—Model Development and Deployment 6. Supervised Models Decision Trees Managing Temperature Effects in Edge IoT Deployments and Adaptive Coefficients Temperature Variations and Sensor Data Errors Design of Adaptive System to Autocorrect Model Development (on Paper) Model Development (Python) Model Development (Citizen Data Scientist) Using Machine Learning User Interface Constrained Environment Considerations Decision Rule Build the Model Inference Pre-Run Compute & Memory Statistics Post Run - Compute & Memory Statistics Hardware Economy—Model Accuracy tradeoff Modeling for no connectivity Modeling Low Power Bluetooth Connectivity Modeling Wi-Fi Connectivity Connectivity—Model Accuracy tradeoff Modeling Memory Modeling Processing Power Modeling Storage Modeling Environmental Perturbations Connectivity to Hardware Tradeoff Modeling: Hardware Refresh Modeling Over-The-Air (OTA) Firmware Update Modeling Active Learner vs. Lazy Learner Modeling Model Invocation XGBoost AdaBoost Algorithm Generalization of AdaBoost as Gradient Boosting XGBoost Install Coding XGBoost Code The Output Constrained Environment Considerations Decision Paths—Simplified Rules Time Complexity Output Hardware Economy—Model Accuracy trade-off Modeling No Connectivity Modeling Low Power Bluetooth Connectivity Modeling Wi-Fi Connectivity Connectivity—Model Accuracy tradeoff Modeling Memory Modeling Processing Power Modeling Storage Modeling Environmental Perturbations Connectivity to Hardware trade-off Modeling Hardware Refresh Modeling Over-The-Air (OTA) Firmware Update Modeling Active Learner vs. Lazy Learner Modeling Model Invocation Random Forrest Output Random Forest Model Image Model Generated Code Model Equation Naïve Bayesian Bayesian for Multivariate Multinomial Naiïve Bayes Bernoulli Naive Bayes Gaussian Naive Bayes Device Health Prognostics Data Issues Device Malfunction Observations Data Models Naive Bayesian Model in Python Output Pre-Run Compute & Memory Statistics Post Run - Compute & Memory Statistics Constrained Environment Considerations Naïve Bayesian C Code & ML deployment using device Firmware ML Mode C Constants Embed C Code Time Complexity Output Hardware Economy—Model Accuracy trade-off Modeling no Connectivity Modeling Low Power Bluetooth Connectivity Modeling Wi Fi Connectivity Connectivity—Model Accuracy trade-off Modeling Memory Modeling Processing Power Modeling Storage Modeling Environmental Perturbations Connectivity to Hardware trade-off Modeling: Hardware Refresh Modeling Over-The-Air Firmware (OTA) Update Modeling Active Learner vs. Lazy Learner Modeling Model Invocation Linear Regression Crowdedness to Temperature Modeling (Edge State Model) Dataset Model Development Model Validation F-Test T-Test Data Assumptions Assumptions 2: Linear - Residuals are Independent Model Equation Model Equation & Independent Parameters Coefficients Model Development in Python Checking Linearity Scatter Diagram Constrained Environment Considerations Time Complexity Pre-Run Compute & Memory Statistics Post-Run - Compute & Memory Statistics Output Hardware Economy—Model Accuracy trade-off Modeling no Connectivity Modeling Low Power Bluetooth Connectivity Modeling Wi Fi Connectivity Connectivity—Model Accuracy trade-off Modeling Memory Modeling Processing Power Modeling Storage Modeling Environmental Perturbations Connectivity to Hardware trade-off Modeling: Hardware Refresh Modeling Over-The-Air (OTA) Firmware Update Modeling Active Learner vs. Lazy Learner Modeling Model Invocation Kalman Filter Kalman Filter Block Diagram Representation Kalman Filter for Smart City Constrained Environment Considerations Computational Complexity Output Pre-Run Compute & Memory Statistics Post-Run Compute & Memory Statistics Hardware Economy—Model Accuracy trade-off Modeling no Connectivity Modeling Low Power Bluetooth Connectivity Modeling Wi-Fi Connectivity Connectivity—Model Accuracy trade-off Modeling Memory Modeling Processing Power Modeling Storage Modeling Environmental Perturbations Connectivity to Hardware trade-off Modeling Hardware Refresh Modeling Over-The-Air (OTA) Firmware Update Modeling Active Learner vs. Lazy Learner Modeling Model Invocation References 7. Unsupervised Models Hierarchical Clustering Merging Cluster Techniques Agglomerative Cluster (Python) Code Agglomerative Hierarchical Code in C Single Linkage Distance Formula Time Complexity Pre-Run Compute & Memory Statistics Post-Run Compute & Memory Statistics Hardware Economy—Model Accuracy trade-off Modeling no Connectivity Modeling Low Power Bluetooth Connectivity Modeling Wi-Fi Connectivity Connectivity—Model Accuracy trade-off Modeling Memory Modeling Processing Power Modeling Storage Modeling Environmental Perturbations Connectivity to Hardware Trade-off Modeling Hardware Refresh Modeling Over-The-Air (OTA) Firmware Update Modeling Active Learner vs. Lazy Learner Modeling Model Invocation Deployment of Climate Models in Extremely Constrained Devices (xCDs) Data Attributes Local Temperatures Hyderabad India, Climate Data (From 1850 to 2012) San Francisco Climate Data (From 1850 to 2012) Climate Change—Hierarchical Cluster Clusters to C Array K-Means Clustering Time Complexity of K-Means Use Case: Sensor Signal and Data Interference & Machine Learning K-Means Example K-Means Clustering—Python Code Computational Complexity Output Hardware Economy—Model Accuracy trade-off Modeling no Connectivity Modeling Low Power Bluetooth Connectivity Modeling Wi-Fi Connectivity Connectivity—Model Accuracy tradeoff Modeling Memory Modeling Processing Power Modeling Storage Modeling Environmental Perturbations Connectivity to Hardware trade-off Modeling Hardware Refresh Modeling Over-The-Air (OTA) Firmware Update Modeling Active Learner vs. Lazy Learner Modeling Model Invocation References Section IV—Democratization & Future of AI 8. National Strategies National Technology Strategies for Serving People Artificial Intelligence of the American People China's New Generation of Artificial Intelligence Development Plan Strategic Goals Japan Strategic Council for AI & Strategy Germany Government—AI Strategy National Institution for Transforming India Aayog national strategy for AI French Strategy for Artificial Intelligence—AI for Humanity The United Nations AI Technology Strategy The Role of the UN AI in the Hands of People References 9. Future Democratization of Artificial Intelligence for the Future of Humanity Appendix Appendix A Windows AI Platform—AI Platform for Windows Developers nVidia Jetson TX2 Google Edge TPU Intel Low Power VPU Neural Compute Engine Arduino Board Specification Milk Producing Data Center using AI Index
Donate to keep this site alive
1. Disable the AdBlock plugin. Otherwise, you may not get any links.
2. Solve the CAPTCHA.
3. Click download link.
4. Lead to download server to download.