Machine Learning: An Algorithmic Perspective, Second ...

1 Chapman & Hall/CRC Machine learning & Pattern Recognition SeriesChapman & Hall/CRC Machine learning & Pattern Recognition SeriesMachine LearningMACHINE LEARNINGAn Algorithmic PerspectiveSecond EditionMarslandStephen Marsland Access online or download to your smartphone, tablet or PC/Mac Search the full text of this and other titles you own Make and share notes and highlights Copy and paste text and figures for use in your own documents Customize your view by changing font size and layoutWITH VITALSOURCE EBOOK Second editionMachine learning : An Algorithmic Perspective, Second Edition helps you understand the algorithms of Machine learning . It puts you on a path toward mastering the relevant mathematics and statistics as well as the necessary programming and to the Second Edition Two new chapters on deep belief networks and Gaussian processes Reorganization of the chapters to make a more natural flow of content Revision of the support vector Machine material, including a simple implementation for experiments New material on random forests, the perceptron convergence theorem, accuracy methods, and conjugate gradient optimization for the multi-layer perceptron Additional discussions of the Kalman and particle filters Improved code, including better use of naming conventions in PythonThe text strongly encourages you to practice with the code.

2 Each chapter includes detailed examples along with further reading and problems. All of the Python code used to create the examples is available on the author s website. Features Reflects recent developments in Machine learning , including the rise of deep belief networks Presents the necessary preliminaries, including basic probability and statistics Discusses supervised learning using neural networks Covers dimensionality reduction , the EM algorithm, nearest neighbor methods, optimal decision boundaries, kernel methods, and optimization Describes evolutionary learning , reinforcement learning , tree-based learners, and methods to combine the predictions of many learners Examines the importance of unsupervised learning , with a focus on the self-organizing feature map Explores modern, statistically based approaches to Machine learning 18/19/14 10:02 AMMACHINE LEARNINGAn Algorithmic PerspectiveSecond 18/26/14 12.

3 45 PMChapman & Hall/CRC Machine learning & Pattern Recognition SeriesSERIES EDITORSRalf HerbrichAmazon Development CenterBerlin, GermanyThore GraepelMicrosoft Research , UKAIMS AND SCOPEThis series reflects the latest advances and applications in Machine learning and pattern recog-nition through the publication of a broad range of reference works, textbooks, and handbooks. The inclusion of concrete examples, applications, and methods is highly encouraged. The scope of the series includes, but is not limited to, titles in the areas of Machine learning , pattern rec-ognition, computational intelligence, robotics, computational/statistical learning theory, natural language processing, computer vision, game AI, game theory, neural networks, computational neuroscience, and other relevant topics, such as Machine learning applied to bioinformatics or cognitive science, which might be proposed by potential TITLESBAYESIAN PROGRAMMING Pierre Bessi re, Emmanuel Mazer, Juan-Manuel Ahuactzin, and Kamel MekhnachaUTILITY-BASED learning FROM DATAC raig Friedman and Sven SandowHANDBOOK OF NATURAL LANGUAGE PROCESSING, Second EDITIONN itin Indurkhya and Fred J.

4 DamerauCOST-SENSITIVE Machine LEARNINGB alaji Krishnapuram, Shipeng Yu, and Bharat RaoCOMPUTATIONAL TRUST MODELS AND Machine learning Xin Liu, Anwitaman Datta, and Ee-Peng LimMULTILINEAR SUBSPACE learning : dimensionality reduction OF MULTIDIMENSIONAL DATA Haiping Lu, Konstantinos N. Plataniotis, and Anastasios N. VenetsanopoulosMACHINE learning : An Algorithmic Perspective, Second EditionStephen MarslandA FIRST COURSE IN Machine LEARNINGS imon Rogers and Mark GirolamiMULTI-LABEL dimensionality reduction Liang Sun, Shuiwang Ji, and Jieping YeENSEMBLE METHODS: FOUNDATIONS AND ALGORITHMS Zhi-Hua 28/26/14 12:45 PMChapman & Hall/CRC Machine learning & Pattern Recognition SeriesMACHINE LEARNINGAn Algorithmic PerspectiveSecond EditionStephen 38/26/14 12:45 PMCRC PressTaylor & Francis Group6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 2015 by Taylor & Francis Group, LLCCRC Press is an imprint of Taylor & Francis Group, an Informa businessNo claim to original Government worksVersion Date: 20140826 International Standard Book Number-13: 978-1-4665-8333-7 (eBook - PDF)This book contains information obtained from authentic and highly regarded sources.

5 Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the valid-ity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future as permitted under Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or uti-lized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopy-ing, microfilming, and recording, or in any information storage or retrieval system, without written permission from the permission to photocopy or use material electronically from this work, please access ( ) or contact the Copyright Clearance Center, Inc.

6 (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to the Taylor & Francis Web site the CRC Press Web site , for MonikaContentsPrologue to 2nd EditionxviiPrologue to 1st EditionxixCHAPTER1 IF DATA HAD MASS, THE EARTH WOULD BE A BLACK Machine TYPES OF Machine SUPERVISED THE Machine learning A NOTE ON A ROADMAP TO THE BOOK12 FURTHER READING13 CHAPTER2 SOME Weight The Curse of KNOWING WHAT YOU KNOW: TESTING Machine learning Training, Testing, and Validation The Confusion Accuracy The Receiver Operator Characteristic (ROC) Unbalanced Measurement TURNING DATA INTO Minimising Risk30viiviii The Na ve Bayes SOME BASIC Variance and The THE BIAS-VARIANCE TRADEOFF35 FURTHER READING36 PRACTICE QUESTIONS37 CHAPTER3 Neurons, Neural Networks, and Linear THE BRAIN AND THE Hebb s McCulloch and Pitts Limitations of the McCulloch and Pitts Neuronal NEURAL THE The learning Rate The Bias The Perceptron learning An Example of Perceptron learning : Logic LINEAR The Perceptron Convergence The Exclusive Or (XOR) A Useful Another Example: The Pima Indian Preprocessing: Data LINEAR Linear Regression Examples66 FURTHER READING67 PRACTICE QUESTIONS68 CHAPTER4 The Multi-layer GOING GOING BACKWARDS.

7 BACK-PROPAGATION OF The Multi-layer Perceptron Initialising the Different Output Activation Functions81 Contents Sequential and Batch Local Picking Up Minibatches and Stochastic Gradient Other THE MULTI-LAYER PERCEPTRON IN Amount of Training Number of Hidden When to Stop EXAMPLES OF USING THE A Regression Classification with the A Classification Example: The Iris Time-Series Data Compression: The Auto-Associative A RECIPE FOR USING THE DERIVING The Network Output and the The Error of the Requirements of an Activation Back-Propagation of The Output Activation An Alternative Error Function108 FURTHER READING108 PRACTICE QUESTIONS109 CHAPTER5 Radial Basis Functions and RECEPTIVE THE RADIAL BASIS FUNCTION (RBF) Training the RBF INTERPOLATION AND BASIS Bases and Basis The Cubic Fitting the Spline to the Smoothing Higher Beyond the Bounds127 FURTHER READING127 PRACTICE QUESTIONS128x ContentsCHAPTER6 dimensionality LINEAR DISCRIMINANT ANALYSIS (LDA) PRINCIPAL COMPONENTS ANALYSIS (PCA) Relation with the Multi-layer Kernel FACTOR INDEPENDENT COMPONENTS ANALYSIS (ICA) LOCALLY LINEAR Multi-Dimensional Scaling (MDS)

8 147 FURTHER READING150 PRACTICE QUESTIONS151 CHAPTER7 Probabilistic GAUSSIAN MIXTURE The Expectation-Maximisation (EM) Information NEAREST NEIGHBOUR Nearest Neighbour Efficient Distance Computations: the Distance Measures165 FURTHER READING167 PRACTICE QUESTIONS168 CHAPTER8 Support Vector OPTIMAL The Margin and Support A Constrained Optimisation Slack Variables for Non-Linearly Separable Choosing Example: THE SUPPORT VECTOR Machine EXTENSIONS TO THE Multi-Class SVM Regression186 Contents Other Advances187 FURTHER READING187 PRACTICE QUESTIONS188 CHAPTER9 Optimisation and GOING Taylor LEAST-SQUARES The Levenberg Marquardt CONJUGATE Conjugate Gradients Conjugate Gradients and the SEARCH: THREE BASIC Exhaustive Greedy Hill EXPLOITATION AND SIMULATED Comparison208 FURTHER READING209 PRACTICE QUESTIONS209 CHAPTER10 Evolutionary THE GENETIC ALGORITHM (GA) String Evaluating Generating Offspring: Parent GENERATING OFFSPRING: GENETIC Elitism, Tournaments, and USING GENETIC Map Punctuated Example: The Knapsack Example: The Four Peaks Limitations of the Training Neural Networks with Genetic Algorithms225xii GENETIC COMBINING SAMPLING WITH EVOLUTIONARY LEARNING227 FURTHER READING228 PRACTICE QUESTIONS229 CHAPTER11 Reinforcement EXAMPLE: GETTING State and Action Carrots and Sticks: The Reward Action MARKOV DECISION The Markov Probabilities in Markov Decision BACK ON HOLIDAY.

9 USING REINFORCEMENT THE DIFFERENCE BETWEEN SARSA AND USES OF REINFORCEMENT LEARNING246 FURTHER READING247 PRACTICE QUESTIONS247 CHAPTER12 learning with USING DECISION CONSTRUCTING DECISION Quick Aside: Entropy in Information Implementing Trees and Graphs in Implementation of the Decision Dealing with Continuous Computational CLASSIFICATION AND REGRESSION TREES (CART) Gini Regression in CLASSIFICATION EXAMPLE261 FURTHER READING263 PRACTICE QUESTIONS264 Contents xiiiCHAPTER13 Decision by Committee: Ensemble RANDOM Comparison with DIFFERENT WAYS TO COMBINE CLASSIFIERS277 FURTHER READING279 PRACTICE QUESTIONS280 CHAPTER14 Unsupervised THEK-MEANS Dealing with Thek-Means Neural A Better Weight Update Example: The Iris Dataset Using Competitive learning for VECTOR THE SELF-ORGANISING FEATURE The SOM Neighbourhood Network dimensionality and Boundary Examples of Using the SOM300 FURTHER READING300 PRACTICE QUESTIONS303 CHAPTER15 Markov Chain Monte Carlo (MCMC) Random Gaussian Random MONTE CARLO OR THE PROPOSAL MARKOV CHAIN MONTE Markov Chains313xiv The Metropolis Hastings Simulated Annealing (Again) Gibbs Sampling318 FURTHER READING319 PRACTICE QUESTIONS320 CHAPTER16 Graphical BAYESIAN Example: Exam Approximate Making Bayesian MARKOV RANDOM HIDDEN MARKOV MODELS (HMMS) The Forward The Viterbi The Baum Welch or Forward Backward TRACKING The Kalman The Particle Filter350 FURTHER READING355 PRACTICE QUESTIONS356 CHAPTER17 Symmetric Weights and Deep Belief ENERGETIC learning .

10 THE HOPFIELD Associative Making an Associative An Energy Capacity of the Hopfield The Continuous Hopfield STOCHASTIC NEURONS THE BOLTZMANN The Restricted Boltzmann Deriving the CD Supervised The RBM as a Directed Belief DEEP Deep Belief Networks (DBN)388 FURTHER READING393 PRACTICE QUESTIONS393 Contents xvCHAPTER18 Gaussian GAUSSIAN PROCESS Adding learning the Choosing a (set of) Covariance GAUSSIAN PROCESS The Laplace Computing the Implementation410 FURTHER READING412 PRACTICE QUESTIONS413 APPENDIXA INSTALLING PYTHON AND OTHER GETTING Python for MATLAB and R CODE Writing and Importing Control USING NUMPY AND Random Linear One Thing to Be Aware of429 FURTHER READING430 PRACTICE QUESTIONS430 Index431 Prologue to 2nd EditionThere have been some interesting developments in Machine learning over the past four years,since the 1st edition of this book came out.