Artificial Intelligence for Engineers: Basics and Implementations (AI)

Preface
Acknowledgments
Contents
1 Preparation Knowledge: Basics of AI
	1.1 Overview
	1.2 Introduction to Artificial Intelligence
		1.2.1 Why Look into AI?
		1.2.2 What Is AI?
		1.2.3 History of AI
			Spawning (1930–1952)
			Birth (1952 and 1956)
			Symbolic AI (1956–1974)
			First AI Winter (1974–1980)
			Expert System and Connectionism Bloom (1980–1987)
			Second AI Winter (1987–1993)
			Recovery (1993–2011)
			Deep Learning and Big Data Rise (2011–present)
		1.2.4 AI Versus Traditional Engineering Methods
			Practice: Prediction of Object Flying Trajectory (Physics Methods Versus Data Method)
		1.2.5 AI Applications
			AI Applications in All Sectors
			AI Applications in Engineering
	1.3 Basics of AI
		1.3.1 Basic Concepts
			Key Machine Learning Elements
			Data Format
			Machine Learning Workflow
		1.3.2 Common Algorithms
			Overview and Machine Learning Tasks
			Supervised Learning
			Unsupervised Learning
			Reinforcement Learning
			Semi-supervised Learning
			Summary
		1.3.3 Challenges and Issues in Machine Learning
			Data Issues
			Inductive Bias
			Underfitting and Overfitting
	1.4 Practice: Gain First Experience with AI via a Machine Learning Task
2 Tools for Artificial Intelligence
	2.1 Overview of Tools for AI
	2.2 Python
		2.2.1 Introduction to Python Coding Environment
		2.2.2 Basics
		2.2.3 Variables and Data Types
		2.2.4 Operators
		2.2.5 Conditional Control Statements
		2.2.6 Sequential Control Statements
		2.2.7 Functions
		2.2.8 Input and Output
		2.2.9 Advanced Python Functionality
	2.3 Data Manipulation and Visualization
		2.3.1 NumPy
			NumPy Array
			Array Constructions
			Array Operations
		2.3.2 Pandas
			From NumPy to Pandas
			Series
			Dataframe
		2.3.3 Matplotlib
			Pyplot: Procedural Plotting Interface
			Object-Oriented Plotting Interface
	2.4 General Machine Learning
		2.4.1 Scikit-Learn
			Data Import
			Data Preprocessing
			Using Models
			Saving Models
	2.5 Deep Learning
		2.5.1 Deep Learning Frameworks
		2.5.2 TensorFlow
			Overview of APIs
			Computational Graph
			Variables
			Placeholders and Comprehensive Example
			Comprehensive Example
		2.5.3 Keras
			Installation and Data Preparation
			Model Establishment with the Sequential API
			Model Establishment with the Functional API
			Training and Result Visualization
	2.6 Reinforcement Learning
		2.6.1 Overview of RL Tools
		2.6.2 OpenAI Gym
	2.7 Practice: Use, Compare, and Understand TensorFlow and Keras for Problem-Solving
3 Linear Models
	3.1 Overview
	3.2 Basics of Linear Models
		3.2.1 Simple Explanation of Linear Models
		3.2.2 General Formulation of Basic Linear Model
	3.3 Other Linear Regression Algorithms
		3.3.1 Ridge
		3.3.2 Lasso
	3.4 Logistic Regression for Classification
		3.4.1 Binary Classification
		3.4.2 Multiclass Classification
	3.5 Making Linear Models Nonlinear via Kernel Functions
		3.5.1 Mapping Data to Higher-Dimensional Space with Stretching Functions
		3.5.2 Kernel Functions
	3.6 Practice: Develop Code to Implement the Basic Linear Model
4 Decision Trees
	4.1 Overview
	4.2 Basics of Decision Trees
	4.3 Classic Decision Tree Algorithms
		4.3.1 ID3 Algorithm
		4.3.2 C4.5 Algorithm
		4.3.3 CART Algorithm
		4.3.4 Implementation
	4.4 Issues and Techniques: Overfitting and Pruning
		4.4.1 Pre-Pruning
		4.4.2 Post-Pruning
			Cost-Complexity Pruning (CCP)
			Reduced Error Pruning (REP)
			Pessimistic Error Pruning (PEP)
			Minimum Error Pruning (MEP)
			Comparison and Summary
	4.5 Practice: Decision Trees in Scikit-learn—Training, Tree Plot, and Testing
5 Support Vector Machines
	5.1 Overview
	5.2 Basics of SVM: Hard-Margin SVM
		5.2.1 Basic Formulation
		5.2.2 Dual Formulation
	5.3 Generalization of SVM: Kernel Methods
	5.4 Soft-Margin SVM
		5.4.1 Basic Formulation
		5.4.2 Dual Formulation
	5.5 More About SVM
		5.5.1 SMO Algorithm
		5.5.2 SVM for Multiclass Classification and Regression
	5.6 Practice: Use of SVMs in Scikit-Learn for Classification and Regression
6 Bayesian Algorithms
	6.1 Overview
	6.2 Statistics Background for Machine Learning
		6.2.1 Statistics and Machine Learning
		6.2.2 Frequentists and Bayesians
		6.2.3 Overview of Statistical Inference
		6.2.4 Maximum Likelihood Estimation (MLE)
		6.2.5 Bayesian Estimation
	6.3 Parametric Bayesian Methods
		6.3.1 Naive Bayes Classifier
		6.3.2 Semi-Naive Bayesian Classifier
			One-Dependent Estimator (ODE)
			Variations of ODE
			Tree Augmented Naive Bayes (TAN)
		6.3.3 Bayesian Network
			Structure
			Implementation
	6.4 Bayesian Nonparametrics
		6.4.1 Parametric Versus Nonparametric Models
			Overview
			Parametric Models
			Nonparametric Models
			From Parametric to Nonparametric Bayesian Algorithms
		6.4.2 Gaussian Processes
			Introduction to Gaussian Process
			Modeling Functions Using Multivariate Gaussian
			Making Predictions Using a Prior and Observations
			Example
			Summary
	6.5 Practice: Code Gaussian Naive Bayes Classifier, Try Bayesian Network, and Apply Gaussian Process
7 Artificial Neural Networks
	7.1 Overview
	7.2 Basics of Artificial Neural Networks
		7.2.1 From Biological Neural Network to ANN
		7.2.2 Activation Function
		7.2.3 Perceptron
		7.2.4 Multiple-Layer Feedforward Neural Network
	7.3 Training with Backpropagation
		7.3.1 Concepts
		7.3.2 Backpropagation in a 3-Layer Network
		7.3.3 Backpropagation in Neural Networks with 3+ Layers
	7.4 Implementation
		7.4.1 Practical Skills
		7.4.2 Procedure for An Example
		7.4.3 *Shape and Arrangement of Arrays for Data
	7.5 Other ANN Issues
	7.6 Practice: Modify and Assess the Architecture of an ANN
8 Deep Learning
	8.1 From Artificial Neural Networks to Deep Learning
		8.1.1 Overview
		8.1.2 The First Wave
		8.1.3 The Second Wave
		8.1.4 The Third Wave
		8.1.5 Summary of Enabling Innovations
	8.2 Convolutional Neural Network
		8.2.1 Convolution
			Forward Pass
			Backward Pass
			Padding and Stride
		8.2.2 ReLU
		8.2.3 Pooling
	8.3 Recurrent Neural Network
		8.3.1 Forward Pass
		8.3.2 Backward Pass
	8.4 Practical Deep Learning Skills
		8.4.1 Initialization
			Overview
			Xavier Initialization
			He Initialization
			LeCun Initialization
			Batch Normalization
		8.4.2 Optimization Methods
			SGD
			Momentum
			Nesterov
			AdaGrad
			AdaDelta
			RMSprop
			Adam
			Nadam
		8.4.3 Data Preprocessing and Augmentation
	8.5 Practice: Build AlexNet Using Keras to Address MNIST Image Classification
9 Ensemble Learning
	9.1 Overview
	9.2 Basics of Ensemble Learning
		9.2.1 Definition
		9.2.2 Basic Questions
		9.2.3 Categories of Ensemble Learning Methods
		9.2.4 Essence of Ensemble Learning
		9.2.5 History and Challenge
	9.3 Bagging
		9.3.1 Basic Bagging
		9.3.2 Random Forest
	9.4 Boosting
		9.4.1 AdaBoost
			Loss Function
			Update on Model Weights
			Update on Sample Weights/Distribution
			Pseudo-Code
		9.4.2 Gradient Boosting
	9.5 Stacking
	9.6 Practice: Code and Evaluate Ensemble Learning Methods
10 Clustering
	10.1 Overview
	10.2 Basics of Unsupervised Learning
		10.2.1 From Supervised Learning to Unsupervised Learning
		10.2.2 Framework for Unsupervised Learning
		10.2.3 Overview of Clustering
	10.3 K-Means Clustering
		10.3.1 Math Framework of K-Means Algorithm
		10.3.2 Implementation of K-Means
		10.3.3 Initialization
		10.3.4 Selection of K
		10.3.5 Pros and Cons
	10.4 Mean-Shift Clustering Algorithm
		10.4.1 Pros and Cons
	10.5 Density-Based Spatial Clustering (DBScan)
		10.5.1 Pros and Cons
	10.6 Gaussian Mixture Models (GMM)
		10.6.1 Pros and Cons
	10.7 Hierarchical Agglomerative Clustering (HAC)
		10.7.1 Pros and Cons
	10.8 Evaluation of Clustering
		10.8.1 Overview of Evaluation Metrics
		10.8.2 Internal Evaluation
			Silhouette Coefficient
			Davies-Bouldin Index
			Dunn Index
		10.8.3 External Evaluation
			Rand Index
			Adjusted Rand Index
			Normalized Mutual Information (NMI)
			Fowlkes-Mallows Index
			Contingency Matrix
	10.9 Practice: Test and Modify Clustering Code for Problem-Solving
11 Dimension Reduction
	11.1 Overview
	11.2 Basics of Dimension Reduction
		11.2.1 Concepts and Needs
		11.2.2 Popular Methods and Classification
	11.3 Common Feature Selection Methods
	11.4 Feature Extraction Method 1: Principal Component Analysis
		11.4.1 Concept and Main Idea
		11.4.2 Theoretical Basis
			Deduction Based on Minimum Distance
			Deduction Based on Maximum Variance
		11.4.3 Implementation
	11.5 Feature Extraction Method 2: Linear Discriminant Analysis
		11.5.1 Concept and Main Idea
		11.5.2 Theoretical Basis
			Rayleigh Quotient and Generalized Rayleigh Quotient
			Binary Classification
			Multiclass Classification
		11.5.3 Implementation
	11.6 Practice: Develop and Modify Code for PCA and LDA
12 Anomaly Detection
	12.1 Overview
	12.2 Basics of Anomaly Detection
	12.3 Statistics-Based Methods
		12.3.1 3 Sigma
		12.3.2 Z-Score
		12.3.3 Boxplot
		12.3.4 Grubbs Hypothesis Test
	12.4 Supervised Learning Methods
		12.4.1 Why Not Use Binary Classification for Anomaly Detection?
		12.4.2 Modification of Supervised Classification Methods for Anomaly Detection
	12.5 Unsupervised Machine Learning Methods
		12.5.1 Overview
		12.5.2 Probabilistic Distribution Based: HBOS
		12.5.3 Distance Based: KNN
		12.5.4 Density Based: LOF, COF, INFLO, and LoOP
			Local Outlier Factor (LOF)
			Connectivity-Based Outlier Factor (COF)
			Influenced Outlierness (INFLO)
			Local Outlier Probability (LoOP)
		12.5.5 Clustering Based
		12.5.6 Tree Based
			iForest
			SCiForest
			RRCF
			Pros and Cons
	12.6 Semisupervised Learning Methods
		12.6.1 Overview
		12.6.2 Autoencoder
			Introduction
			Preparation: Packages and Data
			Model Establishment
			Training and Prediction
			Result Evaluation and Visualization
	12.7 Anomaly Detection Issues
		12.7.1 Data Quality
		12.7.2 Imbalanced Distributions
		12.7.3 High-Dimensional Data
		12.7.4 Model Sensitivity
	12.8 Practice: Implement Typical Anomaly Detection Methods
13 Association Rule Learning
	13.1 Overview
	13.2 Basics of Association Rule Learning
		13.2.1 Definition
		13.2.2 Relationships with Other Machine Learning Topics
		13.2.3 Understanding via History
	13.3 Essential Concepts of Association Rules
		13.3.1 Items, Itemsets, and Rules
		13.3.2 Support, Confidence, and Lift
		13.3.3 Association Rule Analysis Using the Concepts
	13.4 Apriori
		13.4.1 Procedure
		13.4.2 Implementation with an Example
		13.4.3 Pros and Cons
	13.5 FP Growth
		13.5.1 Procedure
		13.5.2 Item Header Table
		13.5.3 FP Tree
		13.5.4 Mining FP Tree for Frequent Itemsets
	13.6 Eclat
		13.6.1 Procedure
		13.6.2 Implementation
	13.7 Practice: Perform Association Rule Learning with Eclat
14 Value-Based Reinforcement Learning
	14.1 Overview
	14.2 Basics of Reinforcement Learning
		14.2.1 Basic Concepts
		14.2.2 Markov Decision Process
		14.2.3 Policy Function, State Function, State-Action Function, and Reward Function
		14.2.4 Implementation of RL Environment
			Implementation with OpenAI Gym
			Implementation from Scratch
	14.3 Bellman Equation
		14.3.1 Formulations of Bellman Equation
		14.3.2 Deduction of Bellman Equation
		14.3.3 Use of Bellman Equation in Reinforcement Learning
	14.4 Value-Based RL
		14.4.1 Overview of RL Algorithms
		14.4.2 Q Learning and Sarsa
		14.4.3 Monte Carlo Method
	14.5 Practice: Solve RL Problem Using Q Learning
15 Policy-Based Reinforcement Learning
	15.1 Overview
	15.2 Policy-Based RL vs. Value-Based RL
	15.3 Basic Concepts
	15.4 Objective Function and Policy Gradient Theorem
		15.4.1 Objective Function
		15.4.2 Policy Gradient Theorem
		15.4.3 Simple Episodic Monte Carlo Implementation of Policy Gradient: REINFORCE V1
		15.4.4 Strategies for Improving Policy Gradient Implementation
	15.5 Policy Function
		15.5.1 Linear Policy Function for Discrete Actions: Formulation 1
		15.5.2 Linear Policy Function for Discrete Actions: Formulation 2
		15.5.3 Policy Function for Continuous Actions
	15.6 Common Policy Gradient Algorithms
		15.6.1 More Objective Function Formulations
		15.6.2 Simple Stepwise Monte Carlo Implementation of Policy Gradient: REINFORCE V2
		15.6.3 Actor-Critic
		15.6.4 Actor-Critic with Baseline
		15.6.5 More Policy Gradient Algorithms
	15.7 Practice: Understand and Modify Policy Gradient Code for Addressing RL Problem
A Appendices
	A.1 Overview
	A.2 Mathematics for Machine Learning
		A.2.1 Statistics
			Random Variables
			Probabilities
			Use of Probability in Machine Learning
			Probability Distributions
		A.2.2 Information Theory
		A.2.3 Array Operations
			Matrix Operations
			General Array Operations
			Array Calculus
	A.3 Optimization
		A.3.1 Gradient-Based Methods
		A.3.2 Newton's Method and Quasi-Newton's Methods
		A.3.3 Conjugate Gradient Methods
		A.3.4 Expectation-Maximization Methods
	A.4 Evaluation Metrics
		A.4.1 Overview and Basics
		A.4.2 Classification: Binary
			Confusion Matrix
			ROC and AUC
			Logarithmic Loss
		A.4.3 Classification: Multiclass
			Indirect Methods
			Confusion Matrix
			Logarithmic Loss
			Kappa Coefficient
			Hinge Loss
		A.4.4 Classification: Multi-Label
			Hamming Distance
			Jaccard Similarity Coefficient
		A.4.5 Regression
			Root Mean Squared Error
			Mean Absolute Error
			Mean Squared Error
			Root Mean Squared Logarithmic Error
			R2 and Adjusted R2
		A.4.6 Clustering
			Inertia and Dunn Index
			Silhouette
			Davies-Bouldin Index
			Calinski-Harabasz Index
			Adjusted Rand Index
			Adjusted Mutual Information
Bibliography
Index