具体描述
Machine learning allows computers to learn and discern patterns without actually being programmed. When Statistical techniques and machine learning are combined together they are a powerful tool for analysing various kinds of data in many computer science/engineering areas including, image processing, speech processing, natural language processing, robot control, as well as in fundamental sciences such as biology, medicine, astronomy, physics, and materials.
Introduction to Statistical Machine Learning provides a general introduction to machine learning that covers a wide range of topics concisely and will help you bridge the gap between theory and practice. Part I discusses the fundamental concepts of statistics and probability that are used in describing machine learning algorithms. Part II and Part III explain the two major approaches of machine learning techniques; generative methods and discriminative methods. While Part III provides an in-depth look at advanced topics that play essential roles in making machine learning algorithms more useful in practice. The accompanying MATLAB/Octave programs provide you with the necessary practical skills needed to accomplish a wide range of data analysis tasks.
作者简介
Masashi Sugiyama
Masashi Sugiyama received the degrees of Bachelor of Engineering, Master of Engineering, and Doctor of Engineering in Computer Science from Tokyo Institute of Technology, Japan in 1997, 1999, and 2001, respectively. In 2001, he was appointed Assistant Professor in the same institute, and he was promoted to Associate Professor in 2003. He moved to the University of Tokyo as Professor in 2014. He received an Alexander von Humboldt Foundation Research Fellowship and researched at Fraunhofer Institute, Berlin, Germany, from 2003 to 2004. In 2006, he received a European Commission Program Erasmus Mundus Scholarship and researched at the University of Edinburgh, Edinburgh, UK. He received the Faculty Award from IBM in 2007 for his contribution to machine learning under non-stationarity, the Nagao Special Researcher Award from the Information Processing Society of Japan in 2011 and the Young Scientists' Prize from the Commendation for Science and Technology by the Minister of Education, Culture, Sports, Science and Technology Japan for his contribution to the density-ratio paradigm of machine learning. His research interests include theories and algorithms of machine learning and data mining, and a wide range of applications such as signal processing, image processing, and robot control.
Affiliations and Expertise
Professor, The University of Tokyo, Japan
目录信息
Preface
Part 1
INTRODUCTION
Chapter 1. Statistical Machine Learning
1.1. Types of Learning
1.2. Examples of Machine Learning Tasks
1.3. Structure of This Textbook
Part 2
STATISTICS AND PROBABILITY
Chapter 2. Random Variables and Probability Distributions
2.1. Mathematical Preliminaries
2.2. Probability
2.3. Random Variable and Probability Distribution
2.4. Properties of Probability Distributions
2.5. Transformation of Random Variables
Chapter 3. Examples of Discrete Probability Distributions
3.1. Discrete Uniform Distribution
3.2. Binomial Distribution
3.3. Hypergeometric Distribution
3.4. Poisson Distribution
3.5. Negative Binomial Distribution
3.6. Geometric Distribution
Chapter 4. Examples of Continuous Probability Distributions
4.1. Continuous Uniform Distribution
4.2. Normal Distribution
4.3. Gamma Distribution, Exponential Distribution, and Chi-Squared Distribution
4.4. Beta Distribution
4.5. Cauchy Distribution and Laplace Distribution
4.6. t-Distribution and F-Distribution
Chapter 5. Multidimensional Probability Distributions
5.1. Joint Probability Distribution
5.2. Conditional Probability Distribution
5.3. Contingency Table
5.4. Bayes’ Theorem
5.5. Covariance and Correlation
5.6. Independence
Chapter 6. Examples of Multidimensional Probability Distributions
6.1. Multinomial Distribution
6.2. Multivariate Normal Distribution
6.3. Dirichlet Distribution
6.4. Wishart Distribution
Chapter 7. Sum of Independent Random Variables
7.1. Convolution
7.2. Reproductive Property
7.3. Law of Large Numbers
7.4. Central Limit Theorem
Chapter 8. Probability Inequalities
8.1. Union Bound
8.2. Inequalities for Probabilities
8.3. Inequalities for Expectation
8.4. Inequalities for the Sum of Independent Random Variables
Chapter 9. Statistical Estimation
9.1. Fundamentals of Statistical Estimation
9.2. Point Estimation
9.3. Interval Estimation
Chapter 10. Hypothesis Testing
10.1. Fundamentals of Hypothesis Testing
10.2. Test for Expectation of Normal Samples
10.3. Neyman-Pearson Lemma
10.4. Test for Contingency Tables
10.5. Test for Difference in Expectations of Normal Samples
10.6. Nonparametric Test for Ranks
10.7. Monte Carlo Test
Part 3
GENERATIVE APPROACH TO STATISTICAL PATTERN RECOGNITION
Chapter 11. Pattern Recognition via Generative Model Estimation
11.1. Formulation of Pattern Recognition
11.2. Statistical Pattern Recognition
11.3. Criteria for Classifier Training
11.4. Generative and Discriminative Approaches
Chapter 12. Maximum Likelihood Estimation
12.1. Definition
12.2. Gaussian Model
12.3. Computing the Class-Posterior Probability
12.4. Fisher’s Linear Discriminant Analysis (FDA)
12.5. Hand-Written Digit Recognition
Chapter 13. Properties of Maximum Likelihood Estimation
13.1. Consistency
13.2. Asymptotic Unbiasedness
13.3. Asymptotic Efficiency
13.4. Asymptotic Normality
13.5. Summary
Chapter 14. Model Selection for Maximum Likelihood Estimation
14.1. Model Selection
14.2. KL Divergence
14.3. AIC
14.4. Cross Validation
14.5. Discussion
Chapter 15. Maximum Likelihood Estimation for Gaussian Mixture Model
15.1. Gaussian Mixture Model
15.2. MLE
15.3. Gradient Ascent Algorithm
15.4. EM Algorithm
Chapter 16. Nonparametric Estimation
16.1. Histogram Method
16.2. Problem Formulation
16.3. KDE
16.4. NNDE
Chapter 17. Bayesian Inference
17.1. Bayesian Predictive Distribution
17.2. Conjugate Prior
17.3. MAP Estimation
17.4. Bayesian Model Selection
Chapter 18. Analytic Approximation of Marginal Likelihood
18.1. Laplace Approximation
18.2. Variational Approximation
Chapter 19. Numerical Approximation of Predictive Distribution
19.1. Monte Carlo Integration
19.2. Importance Sampling
19.3. Sampling Algorithms
Chapter 20. Bayesian Mixture Models
20.1. Gaussian Mixture Models
20.2. Latent Dirichlet Allocation (LDA)
Part 4
DISCRIMINATIVE APPROACH TO STATISTICAL MACHINE LEARNING
Chapter 21. Learning Models
21.1. Linear-in-Parameter Model
21.2. Kernel Model
21.3. Hierarchical Model
Chapter 22. Least Squares Regression
22.1. Method of LS
22.2. Solution for Linear-in-Parameter Model
22.3. Properties of LS Solution
22.4. Learning Algorithm for Large-Scale Data
22.5. Learning Algorithm for Hierarchical Model
Chapter 23. Constrained LS Regression
23.1. Subspace-Constrained LS
23.2. ℓ2-Constrained LS
23.3. Model Selection
Chapter 24. Sparse Regression
24.1. ℓ1-Constrained LS
24.2. Solving ℓ1-Constrained LS
24.3. Feature Selection by Sparse Learning
24.4. Various Extensions
Chapter 25. Robust Regression
25.1. Nonrobustness of ℓ2-Loss Minimization
25.2. ℓ1-Loss Minimization
25.3. Huber Loss Minimization
25.4. Tukey Loss Minimization
Chapter 26. Least Squares Classification
26.1. Classification by LS Regression
26.2. 0∕1-Loss and Margin
26.3. Multiclass Classification
Chapter 27. Support Vector Classification
27.1. Maximum Margin Classification
27.2. Dual Optimization of Support Vector Classification
27.3. Sparseness of Dual Solution
27.4. Nonlinearization by Kernel Trick
27.5. Multiclass Extension
27.6. Loss Minimization View
Chapter 28. Probabilistic Classification
28.1. Logistic Regression
28.2. LS Probabilistic Classification
Chapter 29. Structured Classification
29.1. Sequence Classification
29.2. Probabilistic Classification for Sequences
29.3. Deterministic Classification for Sequences
Part 5
FURTHER TOPICS
Chapter 30. Ensemble Learning
30.1. Decision Stump Classifier
30.2. Bagging
30.3. Boosting
30.4. General Ensemble Learning
Chapter 31. Online Learning
31.1. Stochastic Gradient Descent
31.2. Passive-Aggressive Learning
31.3. Adaptive Regularization of Weight Vectors (AROW)
Chapter 32. Confidence of Prediction
32.1. Predictive Variance for ℓ2-Regularized LS
32.2. Bootstrap Confidence Estimation
32.3. Applications
Chapter 33. Semisupervised Learning
33.1. Manifold Regularization
33.2. Covariate Shift Adaptation
33.3. Class-balance Change Adaptation
Chapter 34. Multitask Learning
34.1. Task Similarity Regularization
34.2. Multidimensional Function Learning
34.3. Matrix Regularization
Chapter 35. Linear Dimensionality Reduction
35.1. Curse of Dimensionality
35.2. Unsupervised Dimensionality Reduction
35.3. Linear Discriminant Analyses for Classification
35.4. Sufficient Dimensionality Reduction for Regression
35.5. Matrix Imputation
Chapter 36. Nonlinear Dimensionality Reduction
36.1. Dimensionality Reduction with Kernel Trick
36.2. Supervised Dimensionality Reduction with Neural Networks
36.3. Unsupervised Dimensionality Reduction with Autoencoder
36.4. Unsupervised Dimensionality Reduction with Restricted Boltzmann Machine
36.5. Deep Learning
Chapter 37. Clustering
37.1. k-Means Clustering
37.2. Kernel k-Means Clustering
37.3. Spectral Clustering
37.4. Tuning Parameter Selection
Chapter 38. Outlier Detection
38.1. Density Estimation and Local Outlier Factor
38.2. Support Vector Data Description
38.3. Inlier-Based Outlier Detection
Chapter 39. Change Detection
39.1. Distributional Change Detection
39.2. Structural Change Detection
References
Index
· · · · · · (收起)
读后感
用户评价
这本书的章节内容组织逻辑严密得让人称赞，读起来有一种享受数学之美的感觉。它没有急于求成地展示那些花哨的深度学习网络，而是花了大篇幅去夯实基础，比如对回归分析、决策树以及无监督学习如聚类方法的深度剖析。我尤其欣赏作者在讨论偏差-方差权衡（Bias-Variance Tradeoff）时的处理方式，他不仅给出了数学上的定义，还结合实际案例分析了在不同数据集规模和模型复杂度下，如何进行有效的正则化处理，这在很多入门书籍中往往是一笔带过的内容。更妙的是，作者在每一章的末尾都设置了“深入探讨”或者“历史背景”的小节，这些旁支信息极大地丰富了我对机器学习发展脉络的理解，让我不再只是机械地套用公式，而是真正理解了这些方法的提出背景和它们在特定历史阶段解决的问题。这种“知其然更知其所以然”的教学方式，极大地提升了我学习的主动性和乐趣。
这本书的封面设计得非常简洁有力，那种深蓝与白色的搭配，透着一股沉稳和专业的气息。我原本对统计学习这个领域有些望而生畏，总觉得里面充斥着晦涩难懂的公式和理论，但这本书的排版和章节划分却给了我极大的信心。作者显然花费了大量心思来构建一个循序渐进的学习路径。它不是那种堆砌知识点的教科书，而是更像一位耐心的导师，引导着读者从最基础的概率论和线性代数概念开始，慢慢过渡到复杂的模型构建。特别是它在介绍每一个核心算法时，都会配有详尽的直观解释和相应的代码示例，这对于我这种偏爱“动手实践”的学习者来说，简直是如获至宝。我记得有一次为了理解支持向量机的拉格朗日对偶问题，我卡了好几天，但书中对该部分的推导和几何解释，一下子就让我茅塞顿开，那种豁然开朗的感觉，至今难忘。它真的做到了将理论的严谨性与实践的可操作性完美地结合起来。
这本书的写作风格非常成熟、沉稳，带有浓厚的学术严谨性，但又不失温度。它的用词精确，逻辑链条清晰，几乎没有冗余的句子。我特别喜欢它在介绍模型局限性时所采取的坦诚态度，它不会过度神化任何一种算法，而是清晰地指出每种方法在面对特定数据结构时的脆弱之处。这培养了我一种健康的怀疑精神，避免了“算法崇拜”的陷阱。阅读过程中，我发现自己的思维方式也悄然发生了转变，开始更加注重问题的定义、假设的有效性以及结果的可信区间，而不仅仅是追求一个“高分”的模型。这本书更像是一份职业素养的培养手册，它不仅教授了如何“做”机器学习，更重要的是，它指导了如何“思考”机器学习，这对于我未来深入研究和职业发展都将是不可或缺的基石。
作为一本面向初学者的进阶读物，这本书的案例选择和数据驱动的讲解方式值得大书特书。它避免了使用那些陈旧的、缺乏实际意义的“鸢尾花”或“西瓜”数据集，转而采用了多个来自真实工业界和学术研究的前沿数据集进行演示。例如，在讨论降维技术时，它不仅讲解了PCA，还详细对比了t-SNE在可视化高维复杂数据时的优劣和适用场景，并且提供了完整的Python代码环境配置指南。这种与时俱进的内容更新，让这本书的生命力得以延续。更重要的是，它教会了我如何批判性地看待模型结果，而不是盲目相信模型给出的准确率数字。作者强调了模型可解释性的重要性，并引入了一些审视模型决策过程的工具，这对于任何想在实际工作中部署机器学习系统的工程师来说，都是至关重要的软技能培养。
坦白说，市面上关于机器学习的书籍汗牛充栋，大多要么过于偏重理论而让人望而却步，要么过于偏重代码实现而缺乏深入的原理阐述。然而，这本书巧妙地找到了一个完美的平衡点。我个人对贝叶斯方法的理解一直比较模糊，但这本书对概率图模型和马尔可夫链蒙特卡洛（MCMC）方法的讲解，清晰得令人印象深刻。作者运用了一种非常具象化的方式来解释复杂的概率分布采样过程，让我能够清晰地“看到”随机变量是如何在状态空间中移动并收敛到目标分布的。阅读过程中，我甚至感觉自己不是在看一本教材，而是在参与一场精心设计的思维实验。它对随机过程的描述极其到位，使得我对时间序列分析和强化学习中涉及的动态规划概念也有了更坚实的认知基础。它成功地将一个公认的难点领域，转化成了可以被清晰把握的知识体系。
👍👍👍👍👍👍👍👍
👍👍👍👍👍👍👍👍
👍👍👍👍👍👍👍👍
👍👍👍👍👍👍👍👍
👍👍👍👍👍👍👍👍
