Statistical Inference and Machine Learning for Big Data (Springer Series in the Data Sciences)

Springer Series in the Data Sciences focuses primarily on monographs and graduate level textbooks. The target audience includes students and researchers working in and across the fields of mathematics, theoretical computer science, and statistics. Data Analysis and Interpretation is a broad field encompassing some of the fastestgrowing subjects in interdisciplinary statistics, mathematics and computer science. It encompasses a process of inspecting, cleaning, transforming, and modeling data with the goal of discovering useful information, suggesting conclusions, and supporting decision making. Data analysis has multiple facets and approaches, including diverse techniques under a variety of names, in different business, science, and social science domains. Springer Series in the Data Sciences addresses the needs of a broad spectrum of scientists and students who are utilizing quantitative methods in their daily research. The series is broad but structured, including topics within all core areas of the data sciences. The breadth of the series reflects the variation of scholarly projects currently underway in the field of machine learning. Preface Acknowledgments Contents List of Acronyms List of Nomenclatures List of Figures List of Tables I. Introduction to Big Data 1. Examples of Big Data 1.1. Multivariate Data 1.2. Categorical Data 1.3. Environmental Data 1.4. Genetic Data 1.5. Time Series Data 1.6. Ranking Data 1.7. Social Network Data 1.8. Symbolic Data 1.9. Image Data II. Statistical Inference for Big Data 2. Basic Concepts in Probability 2.1. Pearson System of Distributions 2.2. Modes of Convergence 2.3. Multivariate Central Limit Theorem 2.4. Markov Chains 3. Basic Concepts in Statistics 3.1. Parametric Estimation 3.2. Hypothesis Testing 3.3. Classical Bayesian Statistics 4. Multivariate Methods 4.1. Matrix Algebra 4.2. Multivariate Analysis as a Generalization of Univariate Analysis 4.2.1. The General Linear Model 4.2.2. One Sample Problem 4.2.3. Two-Sample Problem 4.3. Structure in Multivariate Data Analysis 4.3.1. Principal Component Analysis 4.3.2. Factor Analysis 4.3.3. Canonical Correlation 4.3.4. Linear Discriminant Analysis 4.3.5. Multidimensional Scaling 4.3.6. Copula Methods 5. Nonparametric Statistics 5.1. Goodness-of-Fit Tests 5.2. Linear Rank Statistics 5.3. U Statistics 5.4. Hoeffding's Combinatorial Central Limit Theorem 5.5. Nonparametric Tests 5.5.1. One-Sample Tests of Location 5.5.2. Confidence Interval for the Median 5.5.3. Wilcoxon Signed Rank Test 5.6. Multi-Sample Tests 5.6.1. Two-Sample Tests for Location 5.6.2. Multi-Sample Test for Location 5.6.3. Tests for Dispersion 5.7. Compatibility 5.8. Tests for Ordered Alternatives 5.9. A Unified Theory of Hypothesis Testing 5.9.1. Umbrella Alternatives 5.9.2. Tests for Trend in Proportions 5.10. Randomized Block Designs 5.11. Density Estimation 5.11.1. Univariate Kernel Density Estimation 5.11.2. The Rank Transform 5.11.3. Multivariate Kernel Density Estimation 5.12. Spatial Data Analysis 5.12.1. Spatial Prediction 5.12.2. Point Poisson Kriging of Areal Data 5.13. Efficiency 5.13.1. Pitman Efficiency 5.13.2. Application of Le Cam's Lemmas 5.14. Permutation Methods 6. Exponential Tilting and Its Applications 6.1. Neyman Smooth Tests 6.2. Smooth Models for Discrete Distributions 6.3. Rejection Sampling 6.4. Tweedie's Formula: Univariate Case 6.5. Tweedie's Formula: Multivariate Case 6.6. The Saddlepoint Approximation and Notions of Information 7. Counting Data Analysis 7.1. Inference for Generalized Linear Models 7.2. Inference for Contingency Tables 7.3. Two-Way Ordered Classifications 7.4. Survival Analysis 7.4.1. Kaplan-Meier Estimator 7.4.2. Modeling Survival Data 8. Time Series Methods 8.1. Classical Methods of Analysis 8.2. State Space Modeling 9. Estimating Equations 9.1. Composite Likelihood 9.2. Empirical Likelihood 9.2.1. Application to One-Sample Ranking Problems 9.2.2. Application to Two-Sample Ranking Problems 10. Symbolic Data Analysis 10.1. Introduction 10.2. Some Examples 10.3. Interval Data 10.3.1. Frequency 10.3.2. Sample Mean and Sample Variance 10.3.3. Realization In SODAS 10.4. Multi-nominal Data 10.4.1. Frequency 10.5. Symbolic Regression 10.5.1. Symbolic Regression for Interval Data 10.5.2. Symbolic Regression for Modal Data 10.5.3. Symbolic Regression in SODAS 10.6. Cluster Analysis 10.7. Factor Analysis 10.8. Factorial Discriminant Analysis 10.9. Application to Parkinson's Disease 10.9.1. Data Processing 10.9.2. Result Analysis 10.9.2.1. Viewer 10.9.2.2. Descriptive Statistics 10.9.2.3. Symbolic Regression Analysis 10.9.2.4. Symbolic Clustering 10.9.2.5. Principal Component Analysis 10.9.3. Comparison with Classical Method 10.10. Application to Cardiovascular Disease Analysis 10.10.1. Results of the Analysis 10.10.2. Comparison with the Classical Method III. Machine Learning for Big Data 11. Tools for Machine Learning 11.1. Regression Models 11.2. Simple Linear Regression 11.2.1. Least Squares Method 11.2.2. Statistical Inference on Regression Coefficients 11.2.3. Verifying the Assumptions on the Error Terms 11.3. Multiple Linear Regression 11.3.1. Multiple Linear Regression Model 11.3.2. Normal Equations 11.3.3. Statistical Inference on Regression Coefficients 11.3.4. Model Fit Evaluation 11.4. Regression in Machine Learning 11.4.1. Optimization for Linear Regression in Machine Learning 11.4.1.1. Gradient Descent 11.4.1.2. Feature Standardization 11.4.1.3. Computing Cost on a Test Set 11.5. Classification Models 11.5.1. Logistic Regression 11.5.1.1. Optimization with Maximal Likelihood for Logistic Regression 11.5.1.2. Statistical Inference 11.5.2. Logistic Regression for Binary Classification 11.5.2.1. Kullback-Leibler Divergence 11.5.3. Logistic Regression with Multiple Response Classes 11.5.4. Regularization for Regression Models in Machine Learning 11.5.4.1. Ridge Regression 11.5.4.2. Lasso Regression 11.5.4.3. The Choice of Regularization Method 11.5.5. Support Vector Machines (SVM) 11.5.5.1. Introduction 11.5.5.2. Finding the Optimal Hyperplane 11.5.5.3. SVM for Nonlinearly Separable Data Sets 11.5.5.4. Illustrating SVM 12. Neural Networks 12.1. Feed-Forward Networks 12.1.1. Motivation 12.1.2. Introduction to Neural Networks 12.1.3. Building a Deep Feed-Forward Network 12.1.4. Learning in Deep Networks 12.1.4.1. Quantitative Model 12.1.4.2. Binary Classification Model 12.1.5. Generalization 12.1.5.1. A Machine Learning Approach to Generalization 12.2. Recurrent Neural Networks 12.2.1. Building a Recurrent Neural Network 12.2.2. Learning in Recurrent Networks 12.2.3. Most Common Design Structures of RNNs 12.2.4. Deep RNN 12.2.5. Bidirectional RNN 12.2.6. Long-Term Dependencies and LSTM RNN 12.2.7. Reduction for Exploding Gradients 12.3. Convolution Neural Networks 12.3.1. Convolution Operator for Arrays 12.3.1.1. Properties of the Convolution Operator 12.3.2. Convolution Layers 12.3.3. Pooling Layers 12.4. Text Analytics 12.4.1. Introduction 12.4.2. General Architecture IV. Computational Methods for Statistical Inference 13. Bayesian Computation Methods 13.1. Data Augmentation Methods 13.2. Metropolis-Hastings Algorithm 13.3. Gibbs Sampling 13.4. EM Algorithm 13.4.1. Application to Ranking 13.4.2. Extension to Several Populations 13.5. Variational Bayesian Methods 13.5.1. Optimization of the Variational Distribution 13.6. Bayesian Nonparametric Methods 13.6.1. Dirichlet Prior 13.6.2. The Poisson-Dirichlet Prior 13.6.3. Simulation of Bayesian Posterior Distributions 13.6.4. Other Applications Bibliography Index This book presents a variety of advanced statistical methods at a level suitable for advanced undergraduate and graduate students as well as for others interested in familiarizing themselves with these important subjects. It proceeds to illustrate these methods in the context of real-life applications in a variety of areas such as genetics, medicine, and environmental problems. The book begins in Part I by outlining various data types and by indicating how these are normally represented graphically and subsequently analyzed. In Part II, the basic tools in probability and statistics are introduced with special reference to symbolic data analysis. The most useful and relevant results pertinent to this book are retained. In Part III, the focus is on the tools of machine learning whereas in Part IV the computational aspects of BIG DATA are presented. This book would serve as a handy desk reference for statistical methods at the undergraduate and graduate level as well as be useful in courses which aim to provide an overview of modern statistics and its applications.-- Provided by publisher This book presents a variety of advanced statistical methods at a level suitable for advanced undergraduate and graduate students as well as for others interested in familiarizing themselves with these important subjects. It proceeds to illustrate these methods in the context of real-life applications in a variety of areas such as genetics, medicine, and environmental problems. The book begins in Part I by outlining various data types and by indicating how these are normally represented graphically and subsequently analyzed. In Part II, the basic tools in probability and statistics are introduced with special reference to symbolic data analysis. The most useful and relevant results pertinent to this book are retained. In Part III, the focus is on the tools of machine learning whereas in Part IV the computational aspects of BIG DATA are presented. This book would serve as a handy desk reference for statistical methods at the undergraduate and graduate level as well as be useful in courses which aim to provide an overview of modern statistics and its applications -- Back cover