The Elements of Statistical Learning:Data Mining, Inference, and Prediction

The Elements of Statistical Learning:Data Mining, Inference, and Prediction
分享
扫描下方二维码分享到微信
打开微信,点击右上角”+“,
使用”扫一扫“即可将网页分享到朋友圈。
出版社: Springer
2008-12
ISBN: 9780387848570
装帧: 精装
开本: 其他
纸张: 其他
  • During the past decade there has been an explosion in computation and information technology. With it have come vast amounts of data in a variety of fields such as medicine, biology, finance, and marketing. The challenge of understanding these data has le Trevor Hastie, Robert Tibshirani, and Jerome Friedman are professors of statistics at Stanford University. They are prominent researchers in this area: Hastie and Tibshirani developed generalized additive models and wrote a popular book of that title. Hastie co-developed much of the statistical modeling software and environment in R/S-PLUS and invented principal curves and surfaces. Tibshirani proposed the lasso and is co-author of the very successful An Introduction to the Bootstrap. Friedman is the co-inventor of many data-mining tools including CART, MARS, projection pursuit and gradient boosting. 1 Introduction
    2 Overview of Supervised Learning
    2.1 Introduction
    2.2 Variable Types and Terminology
    2.3 Two Simple Approaches to Prediction:
    Least Squares and Nearest Neighbors
    2.3.1 Linear Models and Least Squares
    2.3.2 Nearest-Neighbor Methods
    2.3.3 From Least Squares to Nearest Neighbors
    2.4 Statistical Decision Theory
    2.5 Local Methods in High Dimensions
    2.6 Statistical Models, Supervised Learning
    and Function Approximation
    2.6.1 A Statistical Model
    for the Joint Distribution Pr(X, Y )
    2.6.2 Supervised Learning
    2.6.3 Function Approximation
    2.7 Structured Regression Models
    2.7.1 Difficulty of the Problem
    2.8 Classes of Restricted Estimators
    2.8.1 Roughness Penalty and Bayesian Methods
    2.8.2 Kernel Methods and Local Regression
    2.8.3 Basis Functions and Dictionary Methods
    2.9 Model Selection and the Bias–Variance Tradeoff
    Bibliographic Notes
    Exercises
    3 Linear Methods for Regression
    3.1 Introduction
    3.2 Linear Regression Models and Least Squares
    3.2.1 Example: Prostate Cancer
    3.2.2 The Gauss–Markov Theorem
    3.2.3 Multiple Regression
    from Simple Univariate Regression
    3.2.4 Multiple Outputs
    3.3 Subset Selection
    3.3.1 Best-Subset Selection
    3.3.2 Forward- and Backward-Stepwise Selection
    3.3.3 Forward-Stagewise Regression
    3.3.4 Prostate Cancer Data Example (Continued)
    3.4 Shrinkage Methods
    3.4.1 Ridge Regression
    3.4.2 The Lasso
    3.4.3 Discussion: Subset Selection, Ridge Regression
    and the Lasso
    3.4.4 Least Angle Regression
    3.5 Methods Using Derived Input Directions
    3.5.1 Principal Components Regression
    3.5.2 Partial Least Squares
    3.6 Discussion: A Comparison of the Selection
    and Shrinkage Methods
    3.7 Multiple Outcome Shrinkage and Selection
    3.8 More on the Lasso and Related Path Algorithms
    3.8.1 Incremental Forward Stagewise Regression
    3.8.2 Piecewise-Linear Path Algorithms
    3.8.3 The Dantzig Selector
    3.8.4 The Grouped Lasso
    3.8.5 Further Properties of the Lasso
    3.8.6 Pathwise Coordinate Optimization
    3.9 Computational Considerations
    Bibliographic Notes
    Exercises

    4 Linear Methods for Classification
    4.1 Introduction
    4.2 Linear Regression of an Indicator Matrix
    4.3 Linear Discriminant Analysis
    4.3.1 Regularized Discriminant Analysis
    4.3.2 Computations for LDA
    4.3.3 Reduced-Rank Linear Discriminant Analysis
    4.4 Logistic Regression
    4.4.1 Fitting Logistic Regression Models
    4.4.2 Example: South African Heart Disease
    4.4.3 Quadratic Approximations and Inference
    4.4.4 L1 Regularized Logistic Regression
    4.4.5 Logistic Regression or LDA?
    4.5 Separating Hyperplanes
    4.5.1 Rosenblatt’s Perceptron Learning Algorithm .
    4.5.2 Optimal Separating Hyperplanes
    Bibliographic Notes
    Exercises
    5 Basis Expansions and Regularization
    5.1 Introduction
    5.2 Piecewise Polynomials and Splines
    5.2.1 Natural Cubic Splines
    5.2.2 Example: South African Heart Disease (Continued)
    5.2.3 Example: Phoneme Recognition
    5.3 Filtering and Feature Extraction
    5.4 Smoothing Splines
    5.4.1 Degrees of Freedom and Smoother Matrices
    5.5 Automatic Selection of the Smoothing Parameters
    5.5.1 Fixing the Degrees of Freedom
    5.5.2 The Bias–Variance Tradeoff
    5.6 Nonparametric Logistic Regression
    5.7 Multidimensional Splines
    5.8 Regularization and Reproducing Kernel Hilbert Spaces
    5.8.1 Spaces of Functions Generated by Kernels
    5.8.2 Examples of RKHS
    5.9 Wavelet Smoothing
    5.9.1 Wavelet Bases and the Wavelet Transform
    5.9.2 Adaptive Wavelet Filtering
    Bibliographic Notes
    Exercises
    Appendix: Computational Considerations for Splines
    Appendix: B-splines
    Appendix: Computations for Smoothing Splines

    6 Kernel Smoothing Methods
    6.1 One-Dimensional Kernel Smoothers
    6.1.1 Local Linear Regression
    6.1.2 Local Polynomial Regression
    6.2 Selecting the Width of the Kernel
    6.3 Local Regression in IRp
    6.4 Structured Local Regression Models in IRp
    6.4.1 Structured Kernels
    6.4.2 Structured Regression Functions
    6.5 Local Likelihood and Other Models
    6.6 Kernel Density Estimation and Classification
    6.6.1 Kernel Density Estimation
    6.6.2 Kernel Density Classification
    6.6.3 The Naive Bayes Classifier
    6.7 Radial Basis Functions and Kernels
    6.8 Mixture Models for Density Estimation and Classification
    6.9 Computational Considerations
    Bibliographic Notes
    Exercises
    7 Model Assessment and Selection
    7.1 Introduction
    7.2 Bias, Variance and Model Complexity
    7.3 The Bias–Variance Decomposition 223
    7.3.1 Example: Bias–Variance Tradeoff
    7.4 Optimism of the Training Error Rate
    7.5 Estimates of In-Sample Prediction Error
    7.6 The Effective Number of Parameters
    7.7 The Bayesian Approach and BIC
    7.8 Minimum Description Length
    7.9 Vapnik–Chervonenkis Dimension
    7.9.1 Example (Continued)
    7.10 Cross-Validation
    7.10.1 K-Fold Cross-Validation
    7.10.2 The Wrong and Right Way
    to Do Cross-validation
    7.10.3 Does Cross-Validation Really Work?
    7.11 Bootstrap Methods
    7.11.1 Example (Continued)
    7.12 Conditional or Expected Test Error?
    Bibliographic Notes
    Exercises
    8 Model Inference and Averaging
    8.1 Introduction
    8.2 The Bootstrap and Maximum Likelihood Methods
    8.2.1 A Smoothing Example
    8.2.2 Maximum Likelihood Inference
    8.2.3 Bootstrap versus Maximum Likelihood
    8.3 Bayesian Methods
    8.4 Relationship Between the Bootstrap
    and Bayesian Inference
    8.5 The EM Algorithm
    8.5.1 Two-Component Mixture Model
    8.5.2 The EM Algorithm in General
    8.5.3 EM as a Maximization–Maximization Procedure
    8.6 MCMC for Sampling from the Posterior
    8.7 Bagging
    8.7.1 Example: Trees with Simulated Data
    8.8 Model Averaging and Stacking
    8.9 Stochastic Search: Bumping
    Bibliographic Notes
    Exercises
    9 Additive Models, Trees, and Related Methods
    9.1 Generalized Additive Models
    9.1.1 Fitting Additive Models
    9.1.2 Example: Additive Logistic Regression
    9.1.3 Summary
    9.2 Tree-Based Methods
    9.2.1 Background
    9.2.2 Regression Trees
    9.2.3 Classification Trees
    9.2.4 Other Issues
    9.2.5 Spam Example (Continued)
    9.3 PRIM: Bump Hunting
    9.3.1 Spam Example (Continued)
    9.4 MARS: Multivariate Adaptive Regression Splines
    9.4.1 Spam Example (Continued)
    9.4.2 Example (Simulated Data)
    9.4.3 Other Issues
    9.5 Hierarchical Mixtures of Experts
    9.6 Missing Data
    9.7 Computational Considerations
    Bibliographic Notes
    Exercises
    10 Boosting and Additive Trees
    10.1 Boosting Methods
    10.1.1 Outline of This Chapter
    10.2 Boosting Fits an Additive Model
    10.3 Forward Stagewise Additive Modeling
    10.4 Exponential Loss and AdaBoost
    10.5 Why Exponential Loss?
    10.6 Loss Functions and Robustness
    10.7 “Off-the-Shelf” Procedures for Data Mining
    10.8 Example: Spam Data
    10.9 Boosting Trees
    10.10 Numerical Optimization via Gradient Boosting
    10.10.1 Steepest Descent
    10.10.2 Gradient Boosting
    10.10.3 Implementations of Gradient Boosting
    10.11 Right-Sized Trees for Boosting
    10.12 Regularization
    10.12.1 Shrinkage
    10.12.2 Subsampling
    10.13 Interpretation
    10.13.1 Relative Importance of Predictor Variables
    10.13.2 Partial Dependence Plots
    10.14 Illustrations
    10.14.1 California Housing
    10.14.2 New Zealand Fish
    10.14.3 Demographics Data
    Bibliographic Notes
    Exercises
    11 Neural Networks
    11.1 Introduction
    11.2 Projection Pursuit Regression
    11.3 Neural Networks
    11.4 Fitting Neural Networks
    11.5 Some Issues in Training Neural Networks
    11.5.1 Starting Values
    11.5.2 Overfitting
    11.5.3 Scaling of the Inputs
    11.5.4 Number of Hidden Units and Layers
    11.5.5 Multiple Minima
    11.6 Example: Simulated Data
    11.7 Example: ZIP Code Data
    11.8 Discussion
    11.9 Bayesian Neural Nets and the NIPS 2003 Challenge
    11.9.1 Bayes, Boosting and Bagging
    11.9.2 Performance Comparisons
    11.10 Computational Considerations
    Bibliographic Notes
    Exercises
    12 Support Vector Machines and
    Flexible Discriminants
    12.1 Introduction
    12.2 The Support Vector Classifier
    12.2.1 Computing the Support Vector Classifier
    12.2.2 Mixture Example (Continued)
    12.3 Support Vector Machines and Kernels
    12.3.1 Computing the SVM for Classification
    12.3.2 The SVM as a Penalization Method
    12.3.3 Function Estimation and Reproducing Kernels
    12.3.4 SVMs and the Curse of Dimensionality
    12.3.5 A Path Algorithm for the SVM Classifier
    12.3.6 Support Vector Machines for Regression
    12.3.7 Regression and Kernels
    12.3.8 Discussion
    12.4 Generalizing Linear Discriminant Analysis
    12.5 Flexible Discriminant Analysis
    12.5.1 Computing the FDA Estimates
    12.6 Penalized Discriminant Analysis
    12.7 Mixture Discriminant Analysis
    12.7.1 Example: Waveform Data
    Bibliographic Notes
    Exercises
    13 Prototype Methods and Nearest-Neighbors
    13.1 Introduction
    13.2 Prototype Methods
    13.2.1 K-means Clustering
    13.2.2 Learning Vector Quantization
    13.2.3 Gaussian Mixtures
    13.3 k-Nearest-Neighbor Classifiers
    13.3.1 Example: A Comparative Study
    13.3.2 Example: k-Nearest-Neighbors
    and Image Scene Classification
    13.3.3 Invariant Metrics and Tangent Distance
    13.4 Adaptive Nearest-Neighbor Methods
    13.4.1 Example
    13.4.2 Global Dimension Reduction
    for Nearest-Neighbors
    13.5 Computational Considerations
    Bibliographic Notes
    Exercises

    14 Unsupervised  Learning
    14.1 Introduction
    14.2 Association Rules
    14.2.1 Market Basket Analysis
    14.2.2 The Apriori Algorithm
    14.2.3 Example: Market Basket Analysis
    14.2.4 Unsupervised as Supervised Learning
    14.2.5 Generalized Association Rules
    14.2.6 Choice of Supervised Learning Method
    14.2.7 Example: Market Basket Analysis (Continued)
    14.3 Cluster Analysis
    14.3.1 Proximity Matrices
    14.3.2 Dissimilarities Based on Attributes
    14.3.3 Object Dissimilarity
    14.3.4 Clustering Algorithms
    14.3.5 Combinatorial Algorithms
    14.3.6 K-means
    14.3.7 Gaussian Mixtures as Soft K-means Clustering
    14.3.8 Example: Human Tumor Microarray Data
    14.3.9 Vector Quantization
    14.3.10   K-medoids
    14.3.11   Practical Issues
    14.3.12  Hierarchical Clustering
    14.4 Self-Organizing Maps
    14.5 Principal Components, Curves and Surfaces
    14.5.1 Principal Components
    14.5.2 Principal Curves and Surfaces
    14.5.3 Spectral Clustering
    14.5.4 Kernel Principal Components
    14.5.5 Sparse Principal Components
    14.6 Non-negative Matrix Factorization
    14.6.1 Archetypal Analysis
    14.7 Independent Component Analysis
    and Exploratory Projection Pursuit
    14.7.1 Latent Variables and Factor Analysis
    14.7.2 Independent Component Analysis
    14.7.3 Exploratory Projection Pursuit
    14.7.4 A Direct Approach to ICA
    14.8 Multidimensional Scaling
    14.9 Nonlinear Dimension Reduction
    and Local Multidimensional Scaling
    14.10  The Google PageRank Algorithm
    Bibliographic Notes
    Exercises

    15 Random Forests
    15.1 Introduction
    15.2 Definition of Random Forests
    15.3 Details of Random Forests
    15.3.1 Out of Bag Samples
    15.3.2 Variable Importance
    15.3.3 Proximity Plots
    15.3.4 Random Forests and Overfitting
    15.4 Analysis of Random Forests
    15.4.1 Variance and the De-Correlation Effect
    15.4.2 Bias
    15.4.3 Adaptive Nearest Neighbors
    Bibliographic Notes
    Exercises
    16 Ensemble Learning
    16.1 Introduction
    16.2 Boosting and Regularization Paths
    16.2.1 Penalized Regression
    16.2.2 The “Bet on Sparsity” Principle
    16.2.3 Regularization Paths, Over-fitting and Margins
    16.3 Learning Ensembles
    16.3.1 Learning a Good Ensemble
    16.3.2 Rule Ensembles
    Bibliographic Notes
    Exercises
    17 Undirected Graphical Models
    17.1 Introduction
    17.2 Markov Graphs and Their Properties
    17.3 Undirected Graphical Models for Continuous Variables
    17.3.1 Estimation of the Parameters
    when the Graph Structure is Known
    17.3.2 Estimation of the Graph Structure
    17.4 Undirected Graphical Models for Discrete Variables
    17.4.1 Estimation of the Parameters
    when the Graph Structure is Known
    17.4.2 Hidden Nodes
    17.4.3 Estimation of the Graph Structure
    17.4.4 Restricted Boltzmann Machines
    Exercises
    18 High-Dimensional Problems: p ≫ N
    18.1 When p is Much Bigger than N
    18.2 Diagonal Linear Discriminant Analysis
    and Nearest Shrunken  Centroids
    18.3 Linear Classifiers with Quadratic Regularization
    18.3.1 Regularized  Discriminant Analysis
    18.3.2 Logistic Regression
    with Quadratic Regularization
    18.3.3 The Support Vector Classifier
    18.3.4 Feature  Selection
    18.3.5 Computational Shortcuts When p ≫ N
    18.4 Linear Classifiers with L1 Regularization
    18.4.1 Application of Lasso
    to Protein Mass Spectroscopy
    18.4.2 The Fused Lasso for Functional Data
    18.5 Classification When Features are Unavailable
    18.5.1 Example: String Kernels
    and Protein Classification
    18.5.2 Classification and Other Models Using
    Inner-Product Kernels and Pairwise Distances .
    18.5.3 Example: Abstracts Classification
    18.6 High-Dimensional Regression: Supervised Principal Components
    18.6.1 Connection to Latent-Variable Modeling
    18.6.2 Relationship with Partial Least Squares
    18.6.3 Pre-Conditioning for Feature Selection
    18.7 Feature Assessment and the Multiple-Testing Problem
    18.7.1 The False Discovery Rate
    18.7.2 Asymmetric Cutpoints and the SAM Procedure
    18.7.3 A Bayesian Interpretation of the FDR
    18.8 Bibliographic Notes
    Exercises
  • 内容简介:
    During the past decade there has been an explosion in computation and information technology. With it have come vast amounts of data in a variety of fields such as medicine, biology, finance, and marketing. The challenge of understanding these data has le
  • 作者简介:
    Trevor Hastie, Robert Tibshirani, and Jerome Friedman are professors of statistics at Stanford University. They are prominent researchers in this area: Hastie and Tibshirani developed generalized additive models and wrote a popular book of that title. Hastie co-developed much of the statistical modeling software and environment in R/S-PLUS and invented principal curves and surfaces. Tibshirani proposed the lasso and is co-author of the very successful An Introduction to the Bootstrap. Friedman is the co-inventor of many data-mining tools including CART, MARS, projection pursuit and gradient boosting.
  • 目录:
    1 Introduction
    2 Overview of Supervised Learning
    2.1 Introduction
    2.2 Variable Types and Terminology
    2.3 Two Simple Approaches to Prediction:
    Least Squares and Nearest Neighbors
    2.3.1 Linear Models and Least Squares
    2.3.2 Nearest-Neighbor Methods
    2.3.3 From Least Squares to Nearest Neighbors
    2.4 Statistical Decision Theory
    2.5 Local Methods in High Dimensions
    2.6 Statistical Models, Supervised Learning
    and Function Approximation
    2.6.1 A Statistical Model
    for the Joint Distribution Pr(X, Y )
    2.6.2 Supervised Learning
    2.6.3 Function Approximation
    2.7 Structured Regression Models
    2.7.1 Difficulty of the Problem
    2.8 Classes of Restricted Estimators
    2.8.1 Roughness Penalty and Bayesian Methods
    2.8.2 Kernel Methods and Local Regression
    2.8.3 Basis Functions and Dictionary Methods
    2.9 Model Selection and the Bias–Variance Tradeoff
    Bibliographic Notes
    Exercises
    3 Linear Methods for Regression
    3.1 Introduction
    3.2 Linear Regression Models and Least Squares
    3.2.1 Example: Prostate Cancer
    3.2.2 The Gauss–Markov Theorem
    3.2.3 Multiple Regression
    from Simple Univariate Regression
    3.2.4 Multiple Outputs
    3.3 Subset Selection
    3.3.1 Best-Subset Selection
    3.3.2 Forward- and Backward-Stepwise Selection
    3.3.3 Forward-Stagewise Regression
    3.3.4 Prostate Cancer Data Example (Continued)
    3.4 Shrinkage Methods
    3.4.1 Ridge Regression
    3.4.2 The Lasso
    3.4.3 Discussion: Subset Selection, Ridge Regression
    and the Lasso
    3.4.4 Least Angle Regression
    3.5 Methods Using Derived Input Directions
    3.5.1 Principal Components Regression
    3.5.2 Partial Least Squares
    3.6 Discussion: A Comparison of the Selection
    and Shrinkage Methods
    3.7 Multiple Outcome Shrinkage and Selection
    3.8 More on the Lasso and Related Path Algorithms
    3.8.1 Incremental Forward Stagewise Regression
    3.8.2 Piecewise-Linear Path Algorithms
    3.8.3 The Dantzig Selector
    3.8.4 The Grouped Lasso
    3.8.5 Further Properties of the Lasso
    3.8.6 Pathwise Coordinate Optimization
    3.9 Computational Considerations
    Bibliographic Notes
    Exercises

    4 Linear Methods for Classification
    4.1 Introduction
    4.2 Linear Regression of an Indicator Matrix
    4.3 Linear Discriminant Analysis
    4.3.1 Regularized Discriminant Analysis
    4.3.2 Computations for LDA
    4.3.3 Reduced-Rank Linear Discriminant Analysis
    4.4 Logistic Regression
    4.4.1 Fitting Logistic Regression Models
    4.4.2 Example: South African Heart Disease
    4.4.3 Quadratic Approximations and Inference
    4.4.4 L1 Regularized Logistic Regression
    4.4.5 Logistic Regression or LDA?
    4.5 Separating Hyperplanes
    4.5.1 Rosenblatt’s Perceptron Learning Algorithm .
    4.5.2 Optimal Separating Hyperplanes
    Bibliographic Notes
    Exercises
    5 Basis Expansions and Regularization
    5.1 Introduction
    5.2 Piecewise Polynomials and Splines
    5.2.1 Natural Cubic Splines
    5.2.2 Example: South African Heart Disease (Continued)
    5.2.3 Example: Phoneme Recognition
    5.3 Filtering and Feature Extraction
    5.4 Smoothing Splines
    5.4.1 Degrees of Freedom and Smoother Matrices
    5.5 Automatic Selection of the Smoothing Parameters
    5.5.1 Fixing the Degrees of Freedom
    5.5.2 The Bias–Variance Tradeoff
    5.6 Nonparametric Logistic Regression
    5.7 Multidimensional Splines
    5.8 Regularization and Reproducing Kernel Hilbert Spaces
    5.8.1 Spaces of Functions Generated by Kernels
    5.8.2 Examples of RKHS
    5.9 Wavelet Smoothing
    5.9.1 Wavelet Bases and the Wavelet Transform
    5.9.2 Adaptive Wavelet Filtering
    Bibliographic Notes
    Exercises
    Appendix: Computational Considerations for Splines
    Appendix: B-splines
    Appendix: Computations for Smoothing Splines

    6 Kernel Smoothing Methods
    6.1 One-Dimensional Kernel Smoothers
    6.1.1 Local Linear Regression
    6.1.2 Local Polynomial Regression
    6.2 Selecting the Width of the Kernel
    6.3 Local Regression in IRp
    6.4 Structured Local Regression Models in IRp
    6.4.1 Structured Kernels
    6.4.2 Structured Regression Functions
    6.5 Local Likelihood and Other Models
    6.6 Kernel Density Estimation and Classification
    6.6.1 Kernel Density Estimation
    6.6.2 Kernel Density Classification
    6.6.3 The Naive Bayes Classifier
    6.7 Radial Basis Functions and Kernels
    6.8 Mixture Models for Density Estimation and Classification
    6.9 Computational Considerations
    Bibliographic Notes
    Exercises
    7 Model Assessment and Selection
    7.1 Introduction
    7.2 Bias, Variance and Model Complexity
    7.3 The Bias–Variance Decomposition 223
    7.3.1 Example: Bias–Variance Tradeoff
    7.4 Optimism of the Training Error Rate
    7.5 Estimates of In-Sample Prediction Error
    7.6 The Effective Number of Parameters
    7.7 The Bayesian Approach and BIC
    7.8 Minimum Description Length
    7.9 Vapnik–Chervonenkis Dimension
    7.9.1 Example (Continued)
    7.10 Cross-Validation
    7.10.1 K-Fold Cross-Validation
    7.10.2 The Wrong and Right Way
    to Do Cross-validation
    7.10.3 Does Cross-Validation Really Work?
    7.11 Bootstrap Methods
    7.11.1 Example (Continued)
    7.12 Conditional or Expected Test Error?
    Bibliographic Notes
    Exercises
    8 Model Inference and Averaging
    8.1 Introduction
    8.2 The Bootstrap and Maximum Likelihood Methods
    8.2.1 A Smoothing Example
    8.2.2 Maximum Likelihood Inference
    8.2.3 Bootstrap versus Maximum Likelihood
    8.3 Bayesian Methods
    8.4 Relationship Between the Bootstrap
    and Bayesian Inference
    8.5 The EM Algorithm
    8.5.1 Two-Component Mixture Model
    8.5.2 The EM Algorithm in General
    8.5.3 EM as a Maximization–Maximization Procedure
    8.6 MCMC for Sampling from the Posterior
    8.7 Bagging
    8.7.1 Example: Trees with Simulated Data
    8.8 Model Averaging and Stacking
    8.9 Stochastic Search: Bumping
    Bibliographic Notes
    Exercises
    9 Additive Models, Trees, and Related Methods
    9.1 Generalized Additive Models
    9.1.1 Fitting Additive Models
    9.1.2 Example: Additive Logistic Regression
    9.1.3 Summary
    9.2 Tree-Based Methods
    9.2.1 Background
    9.2.2 Regression Trees
    9.2.3 Classification Trees
    9.2.4 Other Issues
    9.2.5 Spam Example (Continued)
    9.3 PRIM: Bump Hunting
    9.3.1 Spam Example (Continued)
    9.4 MARS: Multivariate Adaptive Regression Splines
    9.4.1 Spam Example (Continued)
    9.4.2 Example (Simulated Data)
    9.4.3 Other Issues
    9.5 Hierarchical Mixtures of Experts
    9.6 Missing Data
    9.7 Computational Considerations
    Bibliographic Notes
    Exercises
    10 Boosting and Additive Trees
    10.1 Boosting Methods
    10.1.1 Outline of This Chapter
    10.2 Boosting Fits an Additive Model
    10.3 Forward Stagewise Additive Modeling
    10.4 Exponential Loss and AdaBoost
    10.5 Why Exponential Loss?
    10.6 Loss Functions and Robustness
    10.7 “Off-the-Shelf” Procedures for Data Mining
    10.8 Example: Spam Data
    10.9 Boosting Trees
    10.10 Numerical Optimization via Gradient Boosting
    10.10.1 Steepest Descent
    10.10.2 Gradient Boosting
    10.10.3 Implementations of Gradient Boosting
    10.11 Right-Sized Trees for Boosting
    10.12 Regularization
    10.12.1 Shrinkage
    10.12.2 Subsampling
    10.13 Interpretation
    10.13.1 Relative Importance of Predictor Variables
    10.13.2 Partial Dependence Plots
    10.14 Illustrations
    10.14.1 California Housing
    10.14.2 New Zealand Fish
    10.14.3 Demographics Data
    Bibliographic Notes
    Exercises
    11 Neural Networks
    11.1 Introduction
    11.2 Projection Pursuit Regression
    11.3 Neural Networks
    11.4 Fitting Neural Networks
    11.5 Some Issues in Training Neural Networks
    11.5.1 Starting Values
    11.5.2 Overfitting
    11.5.3 Scaling of the Inputs
    11.5.4 Number of Hidden Units and Layers
    11.5.5 Multiple Minima
    11.6 Example: Simulated Data
    11.7 Example: ZIP Code Data
    11.8 Discussion
    11.9 Bayesian Neural Nets and the NIPS 2003 Challenge
    11.9.1 Bayes, Boosting and Bagging
    11.9.2 Performance Comparisons
    11.10 Computational Considerations
    Bibliographic Notes
    Exercises
    12 Support Vector Machines and
    Flexible Discriminants
    12.1 Introduction
    12.2 The Support Vector Classifier
    12.2.1 Computing the Support Vector Classifier
    12.2.2 Mixture Example (Continued)
    12.3 Support Vector Machines and Kernels
    12.3.1 Computing the SVM for Classification
    12.3.2 The SVM as a Penalization Method
    12.3.3 Function Estimation and Reproducing Kernels
    12.3.4 SVMs and the Curse of Dimensionality
    12.3.5 A Path Algorithm for the SVM Classifier
    12.3.6 Support Vector Machines for Regression
    12.3.7 Regression and Kernels
    12.3.8 Discussion
    12.4 Generalizing Linear Discriminant Analysis
    12.5 Flexible Discriminant Analysis
    12.5.1 Computing the FDA Estimates
    12.6 Penalized Discriminant Analysis
    12.7 Mixture Discriminant Analysis
    12.7.1 Example: Waveform Data
    Bibliographic Notes
    Exercises
    13 Prototype Methods and Nearest-Neighbors
    13.1 Introduction
    13.2 Prototype Methods
    13.2.1 K-means Clustering
    13.2.2 Learning Vector Quantization
    13.2.3 Gaussian Mixtures
    13.3 k-Nearest-Neighbor Classifiers
    13.3.1 Example: A Comparative Study
    13.3.2 Example: k-Nearest-Neighbors
    and Image Scene Classification
    13.3.3 Invariant Metrics and Tangent Distance
    13.4 Adaptive Nearest-Neighbor Methods
    13.4.1 Example
    13.4.2 Global Dimension Reduction
    for Nearest-Neighbors
    13.5 Computational Considerations
    Bibliographic Notes
    Exercises

    14 Unsupervised  Learning
    14.1 Introduction
    14.2 Association Rules
    14.2.1 Market Basket Analysis
    14.2.2 The Apriori Algorithm
    14.2.3 Example: Market Basket Analysis
    14.2.4 Unsupervised as Supervised Learning
    14.2.5 Generalized Association Rules
    14.2.6 Choice of Supervised Learning Method
    14.2.7 Example: Market Basket Analysis (Continued)
    14.3 Cluster Analysis
    14.3.1 Proximity Matrices
    14.3.2 Dissimilarities Based on Attributes
    14.3.3 Object Dissimilarity
    14.3.4 Clustering Algorithms
    14.3.5 Combinatorial Algorithms
    14.3.6 K-means
    14.3.7 Gaussian Mixtures as Soft K-means Clustering
    14.3.8 Example: Human Tumor Microarray Data
    14.3.9 Vector Quantization
    14.3.10   K-medoids
    14.3.11   Practical Issues
    14.3.12  Hierarchical Clustering
    14.4 Self-Organizing Maps
    14.5 Principal Components, Curves and Surfaces
    14.5.1 Principal Components
    14.5.2 Principal Curves and Surfaces
    14.5.3 Spectral Clustering
    14.5.4 Kernel Principal Components
    14.5.5 Sparse Principal Components
    14.6 Non-negative Matrix Factorization
    14.6.1 Archetypal Analysis
    14.7 Independent Component Analysis
    and Exploratory Projection Pursuit
    14.7.1 Latent Variables and Factor Analysis
    14.7.2 Independent Component Analysis
    14.7.3 Exploratory Projection Pursuit
    14.7.4 A Direct Approach to ICA
    14.8 Multidimensional Scaling
    14.9 Nonlinear Dimension Reduction
    and Local Multidimensional Scaling
    14.10  The Google PageRank Algorithm
    Bibliographic Notes
    Exercises

    15 Random Forests
    15.1 Introduction
    15.2 Definition of Random Forests
    15.3 Details of Random Forests
    15.3.1 Out of Bag Samples
    15.3.2 Variable Importance
    15.3.3 Proximity Plots
    15.3.4 Random Forests and Overfitting
    15.4 Analysis of Random Forests
    15.4.1 Variance and the De-Correlation Effect
    15.4.2 Bias
    15.4.3 Adaptive Nearest Neighbors
    Bibliographic Notes
    Exercises
    16 Ensemble Learning
    16.1 Introduction
    16.2 Boosting and Regularization Paths
    16.2.1 Penalized Regression
    16.2.2 The “Bet on Sparsity” Principle
    16.2.3 Regularization Paths, Over-fitting and Margins
    16.3 Learning Ensembles
    16.3.1 Learning a Good Ensemble
    16.3.2 Rule Ensembles
    Bibliographic Notes
    Exercises
    17 Undirected Graphical Models
    17.1 Introduction
    17.2 Markov Graphs and Their Properties
    17.3 Undirected Graphical Models for Continuous Variables
    17.3.1 Estimation of the Parameters
    when the Graph Structure is Known
    17.3.2 Estimation of the Graph Structure
    17.4 Undirected Graphical Models for Discrete Variables
    17.4.1 Estimation of the Parameters
    when the Graph Structure is Known
    17.4.2 Hidden Nodes
    17.4.3 Estimation of the Graph Structure
    17.4.4 Restricted Boltzmann Machines
    Exercises
    18 High-Dimensional Problems: p ≫ N
    18.1 When p is Much Bigger than N
    18.2 Diagonal Linear Discriminant Analysis
    and Nearest Shrunken  Centroids
    18.3 Linear Classifiers with Quadratic Regularization
    18.3.1 Regularized  Discriminant Analysis
    18.3.2 Logistic Regression
    with Quadratic Regularization
    18.3.3 The Support Vector Classifier
    18.3.4 Feature  Selection
    18.3.5 Computational Shortcuts When p ≫ N
    18.4 Linear Classifiers with L1 Regularization
    18.4.1 Application of Lasso
    to Protein Mass Spectroscopy
    18.4.2 The Fused Lasso for Functional Data
    18.5 Classification When Features are Unavailable
    18.5.1 Example: String Kernels
    and Protein Classification
    18.5.2 Classification and Other Models Using
    Inner-Product Kernels and Pairwise Distances .
    18.5.3 Example: Abstracts Classification
    18.6 High-Dimensional Regression: Supervised Principal Components
    18.6.1 Connection to Latent-Variable Modeling
    18.6.2 Relationship with Partial Least Squares
    18.6.3 Pre-Conditioning for Feature Selection
    18.7 Feature Assessment and the Multiple-Testing Problem
    18.7.1 The False Discovery Rate
    18.7.2 Asymmetric Cutpoints and the SAM Procedure
    18.7.3 A Bayesian Interpretation of the FDR
    18.8 Bibliographic Notes
    Exercises
查看详情
您可能感兴趣 / 更多
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Black Book of Buried Secrets
Riordan;Rick
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Technique of parents innovation and independent parents cultivation in sugarcane cross breeding
吴才文
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Hongqiao Story: A Record of Whole-process People’s Democracy Practices in Local Communities
上海市长宁区虹桥街道全过程人民民主基层实践基地 作者;中译语通信息科技(上海)有限公司 译;上海人大全过程人民民主研习实践基地
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Ecological Relations of the Vegetation on the Sand Dunes of Lake Michigan(密歇
Henry Chandler Cowle
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The 14th Five-Year Plan for Vocational Skills Training
中华人民共和国人力资源和社会保障部
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Laws of the People\'s Republic of China (2020)
全国人大常委会法制工作委员会
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Ugly Duckling 丑小鸭 (精装本)—小学英语戏剧绘本
[澳]詹姆斯 · 宾 (澳)吉莉安 · 法拉蒂
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Computer and the Brain 计算机与人脑
John von Neumann约翰·冯
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Last Firehawk 2 :The Crystal Caverns:火鹰传奇
Katrina Charman
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Real Thief
William Steig
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Wizard of Oz 绿野仙踪(精装本)(小学英语戏剧绘本)
[澳]詹姆斯·宾 (澳)吉莉安·法拉蒂
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Pied Piper of Hamelin 花衣魔笛手(精装本)(小学英语戏剧绘本)
[澳]詹姆斯·宾 (澳)吉莉安·法拉蒂
系列丛书 / 更多
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Black Book of Buried Secrets
Riordan;Rick
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Technique of parents innovation and independent parents cultivation in sugarcane cross breeding
吴才文
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Hongqiao Story: A Record of Whole-process People’s Democracy Practices in Local Communities
上海市长宁区虹桥街道全过程人民民主基层实践基地 作者;中译语通信息科技(上海)有限公司 译;上海人大全过程人民民主研习实践基地
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Ecological Relations of the Vegetation on the Sand Dunes of Lake Michigan(密歇
Henry Chandler Cowle
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The 14th Five-Year Plan for Vocational Skills Training
中华人民共和国人力资源和社会保障部
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Laws of the People\'s Republic of China (2020)
全国人大常委会法制工作委员会
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Ugly Duckling 丑小鸭 (精装本)—小学英语戏剧绘本
[澳]詹姆斯 · 宾 (澳)吉莉安 · 法拉蒂
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Computer and the Brain 计算机与人脑
John von Neumann约翰·冯
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Last Firehawk 2 :The Crystal Caverns:火鹰传奇
Katrina Charman
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Real Thief
William Steig
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Wizard of Oz 绿野仙踪(精装本)(小学英语戏剧绘本)
[澳]詹姆斯·宾 (澳)吉莉安·法拉蒂
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Pied Piper of Hamelin 花衣魔笛手(精装本)(小学英语戏剧绘本)
[澳]詹姆斯·宾 (澳)吉莉安·法拉蒂
相关图书 / 更多
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Black Book of Buried Secrets
Riordan;Rick
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Technique of parents innovation and independent parents cultivation in sugarcane cross breeding
吴才文
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Hongqiao Story: A Record of Whole-process People’s Democracy Practices in Local Communities
上海市长宁区虹桥街道全过程人民民主基层实践基地 作者;中译语通信息科技(上海)有限公司 译;上海人大全过程人民民主研习实践基地
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Ecological Relations of the Vegetation on the Sand Dunes of Lake Michigan(密歇
Henry Chandler Cowle
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The 14th Five-Year Plan for Vocational Skills Training
中华人民共和国人力资源和社会保障部
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Laws of the People\'s Republic of China (2020)
全国人大常委会法制工作委员会
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Ugly Duckling 丑小鸭 (精装本)—小学英语戏剧绘本
[澳]詹姆斯 · 宾 (澳)吉莉安 · 法拉蒂
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Computer and the Brain 计算机与人脑
John von Neumann约翰·冯
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Last Firehawk 2 :The Crystal Caverns:火鹰传奇
Katrina Charman
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Real Thief
William Steig
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Wizard of Oz 绿野仙踪(精装本)(小学英语戏剧绘本)
[澳]詹姆斯·宾 (澳)吉莉安·法拉蒂
The Elements of Statistical Learning:Data Mining, Inference, and Prediction
The Pied Piper of Hamelin 花衣魔笛手(精装本)(小学英语戏剧绘本)
[澳]詹姆斯·宾 (澳)吉莉安·法拉蒂