Machine learning for computer scientists and data analysts : from an applied perspective

목차 (Table of Contents)

• CONTENTS
• Part I. Basics of Machine Learning
• 1 What Is Applied Machine Learning? = 3
• 1.1 Introduction = 3
• 1.2 The Machine Learning Pipeline = 5
• 1.3 Knowing the Application and Data = 7
• 1.4 Getting Started Using Python = 13
• 1.5 Metadata Extraction and Data Pre-processing = 15
• 1.6 Data Exploration = 17
• 1.7 A Practice for Performing Exploratory Data Analysis = 18
• 1.7.1 Importing the Required Libraries for EDA = 19
• 1.7.2 Loading the Data Into Dataframe = 19
• 1.7.3 Data Visualization = 20
• 1.7.4 Data Analysis = 27
• 1.7.5 Performance Evaluation Metrics = 30
• 1.8 Putting It All Together = 32
• 1.9 Exercise Problems = 32
• 2 A Brief Review of Probability Theory and Linear Algebra = 35
• 2.1 Introduction = 35
• 2.2 Fundamental of the Probability = 36
• 2.3 Discrete Random Variable = 40
• 2.3.1 Probability Mass Function = 41
• 2.3.2 Cumulative Distribution Function = 44
• 2.3.3 Expectation and Variance = 45
• 2.4 Continuous Random Variable = 50
• 2.4.1 Probability Density Function = 50
• 2.4.2 Expectation and Variance = 53
• 2.5 Common Distributions = 55
• 2.5.1 Discrete Distributions = 56
• 2.5.2 Continuous Distributions = 59
• 2.6 Joint Probability Distributions = 64
• 2.6.1 Joint Distribution : Discrete Random Variables = 64
• 2.6.2 Joint Distribution : Continuous Random Variables = 66
• 2.6.3 Covariance and Correlation = 68
• 2.6.4 Multivariate Gaussian Distribution = 72
• 2.7 Matrix Decomposition = 74
• 2.7.1 Eigenvalue Decomposition = 74
• 2.7.2 Singular Value Decomposition = 76
• 2.8 Putting It All Together = 77
• 2.9 Exercise Problems = 77
• 3 Supervised Learning = 81
• 3.1 Introduction = 81
• 3.2 Preparing Data = 83
• 3.2.1 Data Abstraction = 83
• 3.2.2 Dealing with Missing Data = 87
• 3.2.3 Dealing with Imbalanced Datasets = 90
• 3.3 Regression = 90
• 3.3.1 Linear Regression = 91
• 3.3.2 Multi-Variable Linear Regression = 94
• 3.3.3 Multi-Variable Adaptive Regression Splines (MARS) = 96
• 3.3.4 AutoRegressive Moving Average = 99
• 3.3.5 Bayesian Linear Regression = 101
• 3.3.6 Logistic Regression = 103
• 3.4 Artificial Neural Networks = 105
• 3.4.1 Modeling of Neuron = 105
• 3.4.2 Implementing Logical Gates with ANN = 107
• 3.4.3 Multi-Layer Perceptron = 108
• 3.4.4 Training of MLPs = 111
• 3.4.5 Inference = 114
• 3.4.6 Issues with Multi-Layer Perceptron = 115
• 3.4.7 Instances of Deep Neural Networks = 120
• 3.5 Support Vector Machines = 127
• 3.5.1 SVM Kernels = 130
• 3.5.2 Multiclass Classification = 131
• 3.6 Ensemble Learning = 133
• 3.6.1 Bagging = 133
• 3.6.2 AdaBoost = 135
• 3.6.3 Bootstrap = 137
• 3.6.4 Gradient Boosting = 138
• 3.6.5 Stacking = 139
• 3.7 Other Machine Learning Techniques = 141
• 3.7.1 Bayesian Model Combination = 141
• 3.7.2 Random Forest = 141
• 3.7.3 Tree-Based Methods = 143
• 3.7.4 AutoEncoder = 145
• 3.8 Putting It All Together = 160
• 3.9 Exercise Problems = 161
• 4 Unsupervised Learning = 163
• 4.1 Introduction = 163
• 4.2 Clustering = 163
• 4.2.1 K-Means Clustering = 165
• 4.2.2 Hierarchical Clustering = 170
• 4.2.3 Mixture Models = 175
• 4.3 Unsupervised Neural Networks = 177
• 4.3.1 Self-Organizing Maps = 178
• 4.3.2 Generative Adversarial Networks = 182
• 4.3.3 Deep Belief Nets = 186
• 4.3.4 Method of Moments = 189
• 4.4 Feature Selection Techniques = 191
• 4.4.1 Principal Component Analysis = 191
• 4.4.2 T-Distributed Stochastic Neighbor Embedding = 195
• 4.4.3 Pearson Correlation Coefficient = 201
• 4.4.4 Independent Component Analysis = 203
• 4.4.5 Non-negative Matrix Factorization (NMF) = 206
• 4.5 Multi-Dimensional Scaling = 209
• 4.6 Google Page Ranking Algorithm = 213
• 4.7 Putting It All Together = 214
• 4.8 Exercise Problems = 215
• 5 Reinforcement Learning = 217
• 5.1 Introduction = 217
• 5.2 Q-Learning = 218
• 5.2.1 Accelerated Q-learning by Environment Exploration = 221
• 5.3 TD(λ)-Learning = 223
• 5.4 SARSA Learning = 224
• 5.5 Deep Q-Learning = 225
• 5.6 Policy Optimization = 226
• 5.6.1 Stochastic Policy Gradient = 226
• 5.6.2 REINFORCE = 226
• 5.7 Gradient-Based Policy Optimization = 230
• 5.8 Putting It All Together = 230
• 5.9 Exercise Problems = 231
• Part II. Advanced Machine Learning
• 6 Online Learning = 235
• 6.1 Introduction = 235
• 6.2 Online Supervised Learning = 236
• 6.2.1 First-/Second-Order Online Learning = 236
• 6.2.2 Online Learning with Regularization = 239
• 6.3 Online Unsupervised Learning = 243
• 6.3.1 Online Clustering = 243
• 6.3.2 Other Unsupervised Tasks = 246
• 6.4 Application and Resources = 249
• 6.4.1 Time Series Prediction = 249
• 6.4.2 Information Retrieval = 250
• 6.4.3 Online Portfolio Selection = 251
• 6.4.4 Other Applications : Combined with Deep Learning = 251
• 6.4.5 Resources = 252
• 6.5 Putting It All Together = 253
• 6.6 Exercise Problems = 255
• 7 Recommender Learning = 257
• 7.1 Introduction = 257
• 7.2 The Recommendation Problem = 257
• 7.3 Content-Based Approach = 258
• 7.4 Collaborative Filtering = 260
• 7.4.1 Memory-Based Collaborative Filtering = 261
• 7.4.2 Latent Factor Model = 264
• 7.5 Factorization Machine = 266
• 7.6 Deep Learning Models = 268
• 7.7 Application and Resources = 270
• 7.7.1 Applications = 270
• 7.7.2 Resources = 272
• 7.8 Putting It All Together = 272
• 7.9 Exercise Problems = 274
• 8 Graph Learning = 277
• 8.1 Introduction = 277
• 8.2 Basics of Math = 277
• 8.2.1 Matrix Manipulation = 278
• 8.2.2 Eigendecomposition on Matrix = 278
• 8.2.3 Approximation Theory = 280
• 8.2.4 Graph Representations and Graph Signal = 281
• 8.2.5 Spectral Graph Theory = 283
• 8.3 Graph Neural Network Models = 285
• 8.3.1 Spatial-Based Graph Convolution Networks = 288
• 8.3.2 Spectral-Based Graph Convolution Networks = 296
• 8.3.3 Other Graph Neural Networks = 300
• 8.4 Application and Resources = 300
• 8.5 Put It All Together = 301
• 8.6 Exercise Problems = 302
• 9 Adversarial Machine Learning = 305
• 9.1 Introduction = 305
• 9.2 Adversarial Attacks and Defenses = 307
• 9.2.1 Adversarial Attacks = 307
• 9.2.2 Adversarial Defenses = 317
• 9.3 Experimental Results = 322
• 9.3.1 Network Architecture = 322
• 9.3.2 Performance with Adversarial Attacks = 322
• 9.3.3 Effective Adversarial Training = 324
• 9.4 Putting It All Together = 326
• 9.5 Exercise Problems = 327
• Part III. Machine Learning in the Field
• 10 SensorNet : An Educational Neural Network Framework for
• Low-Power Multimodal Data Classification = 331
• 10.1 Introduction = 331
• 10.2 SensorNet Architecture = 332
• 10.2.1 Deep Neural Networks Overview = 332
• 10.2.2 Signal Preprocessing = 336
• 10.2.3 Neural Network Architecture = 336
• 10.3 SensorNet Evaluation using Three Case Studies = 338
• 10.3.1 Case Study 1 : Physical Activity Monitoring = 339
• 10.3.2 Case Study 2 : Stand-Alone Dual-Mode Tongue
• Drive System (sdTDS) = 341
• 10.3.3 Case Study 3 : Stress Detection = 342
• 10.4 SensorNet Optimization and Complexity Reduction = 343
• 10.4.1 The Number of Convolutional Layers = 344
• 10.4.2 The Number of Filters = 345
• 10.4.3 Filter Shapes = 347
• 10.4.4 Zero-Padding = 348
• 10.4.5 Activation Functions = 349
• 10.5 SensorNet Hardware Architecture Design = 350
• 10.5.1 Exploiting Efficient Parallelism = 351
• 10.5.2 Hardware Performance Parameters = 352
• 10.6 Resources = 353
• 10.7 Exercise Problems = 354
• 11 Transfer Learning in Mobile Health = 359
• 11.1 Introduction = 359
• 11.2 Transfer Learning = 361
• 11.3 Problem Statement = 362
• 11.3.1 Problem Definition = 363
• 11.3.2 Problem Formulation = 363
• 11.4 TransFall Framework Design = 364
• 11.4.1 Vertical Transformation = 365
• 11.4.2 Horizontal Transformation = 368
• 11.4.3 Label Estimation = 370
• 11.5 Validation Approach = 372
• 11.5.1 Overview of the Datasets = 372
• 11.5.2 Cross-Domain Transfer Learning Scenarios = 374
• 11.5.3 Comparison Approach and Performance Metrics = 375
• 11.5.4 Choice of Classification Model = 376
• 11.6 Results = 376
• 11.6.1 Cross-Platform Transfer Learning Results = 376
• 11.6.2 Cross-Subject Transfer Learning Results = 377
• 11.6.3 Hybrid Transfer Learning Results = 378
• 11.6.4 Transformation Module Analysis = 378
• 11.6.5 Parameter Examination = 380
• 11.7 Exercise Problems = 381
• 12 Applied Machine Learning for Computer Architecture Security = 383
• 12.1 Introduction = 383
• 12.1.1 Malware = 383
• 12.1.2 Microarchitectural Side-Channel Attacks = 386
• 12.2 Challenges Associated with Traditional Security Mechanisms = 388
• 12.3 Deployment of Hardware Performance Counters for
• Computer Architecture Security = 389
• 12.4 Application of Machine Learning for Computer
• Architecture Security Countermeasures = 391
• 12.4.1 Feature Selection : Key Microarchitectural Features = 391
• 12.5 ML for Hardware-Assisted Malware Detection :
• Comparative Analysis = 392
• 12.5.1 Experimental Setup and Data Collection = 394
• 12.5.2 Feature Selection and ML Classifiers Implementation = 395
• 12.5.3 Evaluation Results of ML-Based Malware Detectors = 396
• 12.6 ML for Microarchitectural SCAs Detection : Comparative
• Analysis = 398
• 12.6.1 Detection Based on Victim Applications’ HPCs Data = 399
• 12.6.2 ML Classifiers Implementation = 400
• 12.6.3 Evaluation Results of ML-Based SCAs Detectors = 401
• 12.7 Exercise Problems = 403
• 13 Applied Machine Learning for Cloud Resource Management = 405
• 13.1 Introduction = 405
• 13.1.1 Challenge of Diversity = 405
• 13.2 Modern Resource Provisioning Systems : ML Comes to
• the Rescue = 407
• 13.3 Applications of Machine Learning in Resource
• Provisioning Systems = 411
• 13.3.1 Monitoring and Prediction of Applications’ Behavior = 412
• 13.3.2 Using ML for Performance/Cost/Energy Estimation = 414
• 13.3.3 Explore and Optimize the Selection = 417
• 13.3.4 Decision Making = 419
• 13.4 Security Threats in Cloud Rooted from ML-Based RPS = 420
• 13.4.1 Adversarial Machine Learning Attack to RPS = 422
• 13.4.2 Isolation as a Remedy = 424
• 13.5 Exercise Problems = 426
• References = 429
• Index = 449