Muutke küpsiste eelistusi

Introduction to Environmental Data Science [Kõva köide]

(University of British Columbia, Vancouver)
  • Formaat: Hardback, 647 pages, kõrgus x laius x paksus: 250x175x35 mm, kaal: 1330 g
  • Ilmumisaeg: 23-Mar-2023
  • Kirjastus: Cambridge University Press
  • ISBN-10: 1107065550
  • ISBN-13: 9781107065550
Teised raamatud teemal:
Introduction to Environmental Data ScienceSuurem pilt
  • Formaat: Hardback, 647 pages, kõrgus x laius x paksus: 250x175x35 mm, kaal: 1330 g
  • Ilmumisaeg: 23-Mar-2023
  • Kirjastus: Cambridge University Press
  • ISBN-10: 1107065550
  • ISBN-13: 9781107065550
Teised raamatud teemal:
Püsilink: https://www.kriso.ee/db/9781107065550.html
Märksõnad:
"Statistical and machine learning methods have many applications in the environmental sciences, including prediction and data analysis in meteorology, hydrology and oceanography; pattern recognition for satellite images from remote sensing; management ofagriculture and forests; assessment of climate change; and much more. With rapid advances in machine learning in the last decade, this book provides an urgently needed, comprehensive guide to machine learning and statistics for students and researchers interested in environmental data science. It includes intuitive explanations covering the relevant background mathematics, with examples drawn from the environmental sciences. A broad range of topics are covered, including correlation, regression, classification, clustering, neural networks, random forests, boosting, kernel methods, evolutionary algorithms and deep learning, as well as the recent merging of machine learning and physics. End-of-chapter exercises allow readers to develop their problem-solving skills, and online datasets allow readers to practise analysis of real data. William W. Hsieh is a professor emeritus in the Department of Earth, Ocean and Atmospheric Sciences at the University of British Columbia. Known as a pioneer in introducing machine learning to environmental science, he has written over 100 peer-reviewed journal papers on climate variability, machine learning, atmospheric science, oceanography, hydrology and agricultural science. He is the author of the book Machine Learning Methods in the Environmental Sciences (2009, Cambridge University Press), the first single-authored textbook on machine learning for environmental scientists. Currently retired in Victoria, British Columbia, he enjoys growing organic vegetables"--

Arvustused

'As a new wave of machine learning becomes part of our toolbox for environmental science, this book is both a guide to the latest developments and a comprehensive textbook on statistics and data science. Almost everything is covered, from hypothesis testing to convolutional neural networks. The book is enjoyable to read, well explained and economically written, so it will probably become the first place I'll go to read up on any of these topics.' Alan Geer, European Centre for Medium-Range Weather Forecasts (ECMWF) 'William Hsieh has been one of the 'founding fathers' of an exciting new field of using machine learning (ML) in the environmental sciences. His new book provides readers with a solid introduction to the statistical foundation of ML and various ML techniques, as well as with the fundamentals of data science. The unique combination of solid mathematical and statistical backgrounds with modern applications of ML tools in the environmental sciences is an important distinguishing feature of this book. The broad range of topics covered in this book makes it an invaluable reference and guide for researchers and graduate students working in this and related fields.' Vladimir Krasnopolsky, Center for Weather and Climate Prediction, NOAA 'Dr. Hsieh is one of the pioneers of the development of machine learning for the environmental sciences including the development of methods such as nonlinear principal component analysis to provide insights into the ENSO dynamic. Dr. Hsieh has a deep understanding of the foundations of statistics, machine learning, and environmental processes that he is sharing in this timely and comprehensive work with many recent references. It will no doubt become an indispensable reference for our field. I plan to use the book for my graduate environmental forecasting class and recommend the book for a self-guided progression or as a comprehensive reference.' Philippe Tissot, Texas A&M University-Corpus Christi 'There is a need for a forward-looking text on environmental data science and William Hsieh's text succeeds in filling the gap. This comprehensive text covers basic to advanced material ranging from timeless statistical techniques to some of the latest machine learning approaches. His refreshingly engaging style is written to be understood and complemented by a plethora of expressive visuals. Hsieh's treatment of nonlinearity is cutting-edge and the final chapter examines ways to combine machine learning with physics. This text is destined to become a modern classic.' Sue Ellen Haupt, National Center for Atmospheric Research

Muu info

A comprehensive guide to machine learning and statistics for students and researchers of environmental data science.
Preface xv
Notation Used xviii
List of Abbreviations
xix
1 Introduction
1(18)
1.1 Statistics and Machine Learning A
1(5)
1.2 Environmental Data Science A
6(3)
1.3 A Simple Example of Curve Fitting A
9(3)
1.4 Main Types of Data Problems A
12(5)
1.4.1 Supervised Learning A
15(1)
1.4.2 Unsupervised Learning A
16(1)
1.4.3 Reinforced Learning A
17(1)
1.5 Curse of Dimensionality A
17(2)
2 Basics
19(46)
2.1 Data Types A
19(1)
2.2 Probability A
19(3)
2.3 Probability Density A
22(2)
2.4 Expectation and Mean A
24(1)
2.5 Variance and Standard Deviation A
25(1)
2.6 Covariance A
26(1)
2.7 Online Algorithms for Mean, Variance and Covariance C
27(2)
2.8 Median and Median Absolute Deviation A
29(1)
2.9 Quantiles A
30(2)
2.10 Skewness and Kurtosis B
32(1)
2.11 Correlation A
33(8)
2.11.1 Pearson Correlation A
33(3)
2.11.2 Serial Correlation A
36(2)
2.11.3 Spearman Rank Correlation A
38(1)
2.11.4 Kendall Rank Correlation A
39(1)
2.11.5 Biweight Mid correlation B
40(1)
2.12 Exploratory Data Analysis A
41(6)
2.12.1 Histograms A
41(1)
2.12.2 Quantile--Quantile (Q--Q) Plots B
42(3)
2.12.3 Boxplots A
45(2)
2.13 Mahalanobis Distance A
47(2)
2.13.1 Mahalanobis Distance and Principal Component Analysis B
47(2)
2.14 Bayes Theorem A
49(3)
2.15 Classification A
52(2)
2.16 Clustering A
54(2)
2.17 Information Theory B
56(6)
2.17.1 Entropy B
58(1)
2.17.2 Joint Entropy and Conditional Entropy B
59(1)
2.17.3 Relative Entropy B
60(1)
2.17.4 Mutual Information B
61(1)
Exercises
62(3)
3 Probability Distributions
65(36)
3.1 Binomial Distribution A
66(2)
3.2 Poisson Distribution B
68(1)
3.3 Multinomial Distribution B
68(1)
3.4 Gaussian Distribution A
69(4)
3.5 Maximum Likelihood Estimation A
73(2)
3.6 Multivariate Gaussian Distribution B
75(2)
3.7 Conditional and Marginal Gaussian Distributions C
77(1)
3.8 Gamma Distribution B
78(2)
3.9 Beta Distribution B
80(2)
3.10 Von Mises Distribution C
82(1)
3.11 Extreme Value Distributions C
83(3)
3.12 Gaussian Mixture Model B
86(5)
3.12.1 Expectation-Maximization (EM) Algorithm B
90(1)
3.13 Kernel Density Estimation B
91(2)
3.14 Re-expressing Data A
93(2)
3.15 Student t-distribution B
95(2)
3.16 Chi-squared Distribution B
97(2)
Exercises
99(2)
4 Statistical Inference
101(36)
4.1 Hypothesis Testing A
101(3)
4.2 Student t-test A
104(7)
4.2.1 One-Sample t-test A
105(1)
4.2.2 Independent Two-Sample t-test A
105(2)
4.2.3 Dependent t-test for Paired Samples A
107(1)
4.2.4 Serial Correlation A
107(2)
4.2.5 Significance Test for Correlation A
109(2)
4.3 Non-parametric Alternatives to i-test B
111(4)
4.3.1 Wilcoxon--Mann--Whitney Test C
111(3)
4.3.2 Wilcoxon Signed-Rank Test C
114(1)
4.4 Confidence Interval A
115(4)
4.4.1 Confidence Interval for Population Mean B
116(2)
4.4.2 Confidence Interval for Correlation B
118(1)
4.5 Goodness-of-Fit Tests B
119(5)
4.5.1 One-Sample Goodness-of-Fit Tests B
119(2)
4.5.2 Two-Sample Goodness-of-Fit Tests B
121(3)
4.6 Test of Variances B
124(1)
4.7 Mann--Kendall Trend Test B
125(1)
4.8 Bootstrapping A
126(5)
4.9 Field Significance B
131(3)
Exercises
134(3)
5 Linear Regression
137(36)
5.1 Simple Linear Regression A
137(8)
5.1.1 Partition of Sums of Squares A
139(2)
5.1.2 Confidence Interval for Regression Parameters B
141(2)
5.1.3 Confidence Interval and Prediction Interval for the Response Variable B
143(1)
5.1.4 Serial Correlation B
144(1)
5.2 Multiple Linear Regression A
145(8)
5.2.1 Gauss--Markov Theorem B
147(1)
5.2.2 Partition of Sums of Squares B
148(1)
5.2.3 Standardized Predictors A
148(1)
5.2.4 Analysis of Variance (ANOVA) B
149(2)
5.2.5 Confidence and Prediction Intervals B
151(2)
5.3 Multivariate Linear Regression B
153(1)
5.4 Online Learning with Linear Regression C
154(2)
5.5 Circular and Categorical Data A
156(2)
5.6 Predictor Selection C
158(3)
5.7 Ridge Regression B
161(4)
5.8 Lasso C
165(2)
5.9 Quantile Regression C
167(1)
5.10 Generalized Least Squares C
168(2)
5.10.1 Optimal Fingerprinting in Climate Change C
170(1)
Exercises
170(3)
6 Neural Networks
173(43)
6.1 McCulloch and Pitts Model B
174(1)
6.2 Perceptrons A
175(5)
6.2.1 Limitation of Perceptrons B
178(2)
6.3 Multi-layer Perceptrons A
180(9)
6.3.1 Comparison with Polynomials B
185(1)
6.3.2 Hidden Neurons A
186(1)
6.3.3 Monotonic Multi-layer Perception Model C
187(2)
6.4 Extreme Learning Machines A
189(6)
6.4.1 Online Learning B
193(1)
6.4.2 Random Vector Functional Link C
194(1)
6.5 Radial Basis Functions B
195(4)
6.6 Modelling Conditional Distributions B
199(5)
6.6.1 Mixture Density Network B
201(3)
6.7 Quantile Regression C
204(3)
6.8 Historical Development of NN in Environmental Science B
207(7)
6.8.1 Remote Sensing B
208(4)
6.8.2 Hydrology B
212(1)
6.8.3 Atmospheric Science B
213(1)
6.8.4 Oceanography B
213(1)
Exercises
214(2)
7 Non-linear Optimization
216(29)
7.1 Extrema and Saddle Points A
216(3)
7.2 Gradient Vector in Optimization A
219(2)
7.3 Back-Propagation B
221(3)
7.4 Training Protocol A
224(1)
7.5 Gradient Descent Method A
225(2)
7.6 Stochastic Gradient Descent B
227(2)
7.7 Conjugate Gradient Method C
229(3)
7.8 Quasi-Newton Methods C
232(2)
7.9 Non-linear Least Squares Methods C
234(2)
7.10 Evolutionary Algorithms B
236(2)
7.11 Hill Climbing C
238(1)
7.12 Genetic Algorithm C
239(2)
7.13 Differential Evolution C
241(3)
Exercises
244(1)
8 Learning and Generalization
245(38)
8.1 Mean Squared Error and Maximum Likelihood A
245(2)
8.2 Objective Functions and Robustness A
247(3)
8.3 Variance and Bias Errors A
250(1)
8.4 Regularization A
251(4)
8.4.1 Weight Penalty A
251(3)
8.4.2 Early Stopping A
254(1)
8.5 Cross-Validation A
255(3)
8.6 Hyperparameter Tuning A
258(3)
8.7 Ensemble Methods B
261(8)
8.7.1 Bagging A
263(1)
8.7.2 Error of Ensemble B
263(3)
8.7.3 Unequal Weighting of Ensemble Members C
266(2)
8.7.4 Stacking C
268(1)
8.8 Dropout B
269(2)
8.9 Maximum Norm Constraint B
271(1)
8.10 Bayesian Model Selection B
272(1)
8.11 Information Criterion B
273(5)
8.11.1 Bayesian Information Criterion C
273(2)
8.11.2 Akaike Information Criterion C
275(3)
8.12 Bayesian Model Averaging C
278(2)
8.13 Interpretable ML C
280(1)
Exercises
281(2)
9 Principal Components and Canonical Correlation
283(47)
9.1 Principal Component Analysis (PCA) A
283(22)
9.1.1 Geometric Approach to PCA A
284(1)
9.1.2 Eigenvector Approach to PCA A
284(4)
9.1.3 Real and Complex Data C
288(1)
9.1.4 Orthogonality Relations A
289(1)
9.1.5 PCA of the Tropical Pacific Climate Variability A
290(8)
9.1.6 Scaling the PCs and Eigenvectors A
298(1)
9.1.7 Degeneracy of Eigenvalues A
299(1)
9.1.8 A Smaller Covariance Matrix A
299(1)
9.1.9 How Many Modes to Retain B
300(2)
9.1.10 Temporal and Spatial Mean Removal B
302(1)
9.1.11 Singular Value Decomposition B
302(2)
9.1.12 Missing Data C
304(1)
9.2 Rotated PCA B
305(12)
9.2.1 E-frame Rotation B
308(3)
9.2.2 A-frame Rotation B
311(4)
9.2.3 Advantages and Disadvantages of Rotation A
315(2)
9.3 PCA for Two-Dimensional Vectors C
317(3)
9.4 Canonical Correlation Analysis (CCA) B
320(6)
9.4.1 CCA Theory B
320(4)
9.4.2 Pre-filter with PCA B
324(2)
9.5 Maximum Covariance Analysis B
326(1)
Exercises
327(3)
10 Unsupervised Learning
330(42)
10.1 Clustering B
330(7)
10.1.1 Distance Measure B
331(1)
10.1.2 Model Evaluation B
332(5)
10.2 Non-hierarchical Clustering B
337(2)
10.2.1 K-means Clustering B
337(1)
10.2.2 Nucleated Agglomerative Clustering C
338(1)
10.2.3 Gaussian Mixture Model C
339(1)
10.3 Hierarchical Clustering C
339(4)
10.4 Self-Organizing Map C
343(4)
10.4.1 Applications of SOM C
345(2)
10.5 Autoencoder A
347(2)
10.6 Non-linear Principal Component Analysis B
349(14)
10.6.1 Overfitting C
358(2)
10.6.2 Closed Curves C
360(3)
10.7 Other Non-linear Dimensionality Reduction Methods C
363(2)
10.8 Non-linear Canonical Correlation Analysis C
365(4)
Exercises
369(3)
11 Time Series
372(46)
11.1 Fourier Analysis A
372(4)
11.1.1 Fourier Series A
373(1)
11.1.2 Discrete Fourier Transform A
374(1)
11.1.3 Continuous-Time Fourier Transform B
375(1)
11.1.4 Discrete-Time Fourier Transform B
376(1)
11.2 Windows A
376(3)
11.3 Spectrum A
379(14)
11.3.1 Effects of Window Functions A
380(2)
11.3.2 Trend Removal A
382(1)
11.3.3 Nyquist Frequency and Aliasing A
382(2)
11.3.4 Smoothing the Spectrum A
384(1)
11.3.5 Confidence Interval B
385(1)
11.3.6 Examples B
385(1)
11.3.7 Fast Fourier Transform B
386(1)
11.3.8 Relation with Auto-covariance B
387(1)
11.3.9 Rotary Spectrum for 2-D Vectors C
387(3)
11.3.10 Wavelets C
390(3)
11.4 Cross-Spectrum B
393(2)
11.5 Filtering A
395(6)
11.5.1 Periodic Signals A
396(1)
11.5.2 Ideal Filters A
397(1)
11.5.3 Finite Impulse Response Filters B
398(3)
11.6 Averaging A
401(4)
11.6.1 Moving Average Filters A
402(1)
11.6.2 Grid-Scale Noise C
402(2)
11.6.3 Linearization from Time-Averaging B
404(1)
11.7 Singular Spectrum Analysis B
405(5)
11.7.1 Multichannel Singular Spectrum Analysis B
407(3)
11.8 Auto-regressive Process B
410(4)
11.8.1 AR(p) Process B
411(1)
11.8.2 AR(1) Process B
412(1)
11.8.3 AR(2) Process C
413(1)
11.9 Box-Jenkins Models C
414(2)
11.9.1 Moving Average (MA) Process C
414(1)
11.9.2 Auto-regressive Moving Average (ARMA) Model C
414(1)
11.9.3 Auto-regressive Integrated Moving Average (ARIMA) Model C
415(1)
Exercises
416(2)
12 Classification
418(22)
12.1 Linear Discriminant Analysis A
419(5)
12.1.1 Fisher Linear Discriminant B
421(3)
12.2 Logistic Regression A
424(3)
12.2.1 Multiclass Logistic Regression B
425(2)
12.3 Naive Bayes Classifier B
427(1)
12.4 K-nearest Neighbours B
428(2)
12.5 Extreme Learning Machine Classifier A
430(2)
12.6 Cross-Entropy A
432(2)
12.7 Multi-layer Perceptron Classifier A
434(2)
12.8 Class Imbalance A
436(2)
Exercises
438(2)
13 Kernel Methods
440(33)
13.1 From Neural Networks to Kernel Methods B
441(1)
13.2 Primal and Dual Solutions for Linear Regression B
442(2)
13.3 Kernels B
444(4)
13.4 Kernel Ridge Regression B
448(1)
13.5 Advantages and Disadvantages B
449(2)
13.6 Pre-image Problem C
451(2)
13.7 Support Vector Machines (SVM) B
453(10)
13.7.1 Linearly Separable Case B
454(4)
13.7.2 Linearly Non-separable Case B
458(2)
13.7.3 Non-linear Classification by SVM B
460(1)
13.7.4 Multi-class Classification by SVM C
461(1)
13.7.5 Support Vector Regression C
462(1)
13.8 Gaussian Processes B
463(6)
13.8.1 Learning the Hyperparameters B
466(1)
13.8.2 Other Common Kernels C
467(2)
13.9 Kernel Principal Component Analysis C
469(2)
Exercises
471(2)
14 Decision Trees, Random Forests and Boosting
473(21)
14.1 Classification and Regression Trees (CART) A
474(8)
14.1.1 Relative Importance of Predictors B
479(1)
14.1.2 Surrogate Splits B
480(2)
14.2 Random Forests A
482(5)
14.2.1 Extremely Randomized Trees (Extra Trees) B
487(1)
14.3 Boosting A
487(5)
14.3.1 Gradient Boosting B
488(4)
Exercises
492(2)
15 Deep Learning
494(24)
15.1 Transfer Learning A
498(1)
15.2 Convolutional Neural Network B
499(8)
15.2.1 Convolution Operation B
499(3)
15.2.2 Pooling B
502(1)
15.2.3 AlexNet B
503(2)
15.2.4 Residual Neural Network (ResNet) B
505(1)
15.2.5 Data Augmentation C
506(1)
15.2.6 Applications in Environment Science B
506(1)
15.3 Encoder-Decoder Network B
507(3)
15.3.1 U-net B
508(2)
15.4 Time Series C
510(4)
15.4.1 Long Short-Term Memory (LSTM) Network C
510(3)
15.4.2 Temporal Convolutional Network C
513(1)
15.5 Generative Adversarial Network C
514(3)
Exercises
517(1)
16 Forecast Verification and Post-processing
518(31)
16.1 Binary Classes B
519(8)
16.1.1 Skill Scores for Binary Classes B
524(3)
16.2 Multiple Classes C
527(1)
16.3 Probabilistic Forecasts for Binary Classes B
528(3)
16.3.1 Reliability Diagram B
529(2)
16.4 Probabilistic Forecasts for Multiple Classes B
531(1)
16.5 Continuous Variables B
532(3)
16.5.1 Forecast Scores B
532(1)
16.5.2 Skill Scores B
533(2)
16.6 Probabilistic Forecasts for Continuous Variables B
535(1)
16.7 Minimizing Loss C
535(1)
16.8 Spurious Skills A
536(2)
16.9 Extrapolation A
538(3)
16.10 Post-processing C
541(2)
16.11 Downscaling C
543(4)
16.11.1 Reduced Variance C
546(1)
Exercises
547(2)
17 Merging of Machine Learning and Physics
549(20)
17.1 Physics Emulation and Hybrid Models C
550(6)
17.1.1 Radiation in Atmospheric Models C
550(1)
17.1.2 Clouds C
551(2)
17.1.3 Turbulent Fluxes C
553(1)
17.1.4 Hybrid Coupled Atmosphere-Ocean Modelling C
554(1)
17.1.5 Wind Wave Modelling C
555(1)
17.2 Physics-Informed Machine Learning A
556(4)
17.2.1 Soft Constraint C
556(2)
17.2.2 Hard Constraint C
558(2)
17.3 Data Assimilation and ML C
560(8)
17.3.1 3D-Var C
562(1)
17.3.2 4D-Var C
563(1)
17.3.3 Neural Networks in 4D-Var C
564(4)
Exercises
568(1)
Appendices
569(4)
A Trends in Terminology
569(1)
B Lagrange Multipliers
569(4)
References 573(40)
Index 613
William W. Hsieh is a professor emeritus in the Department of Earth, Ocean and Atmospheric Sciences at the University of British Columbia. Known as a pioneer in introducing machine learning to environmental science, he has written over 100 peer-reviewed journal papers on climate variability, machine learning, atmospheric science, oceanography, hydrology, and agricultural science. He is the author of the book Machine Learning Methods in the Environmental Sciences ( Cambridge University Press, 2009), the first single-authored textbook on machine learning for environmental scientists. Currently retired in Victoria, British Columbia, he enjoys growing organic vegetables.