Available:*
Library | Item Barcode | Call Number | Material Type | Item Category 1 | Status |
---|---|---|---|---|---|
Searching... | 30000010177439 | QA76.575 G66 2007 | Open Access Book | Book | Searching... |
Searching... | 30000010184072 | QA76.575 G66 2007 | Open Access Book | Book | Searching... |
On Order
Summary
Summary
Challenges in complexity and variability of multimedia data have led to revolutions in machine learning techniques. Multimedia data, such as digital images, audio streams and motion video programs, exhibit richer structures than simple, isolated data items. A number of pixels in a digital image collectively conveys certain visual content to viewers. A TV video program consists of both audio and image streams that unfold the underlying story. To recognize the visual content of a digital image, or to understand the underlying story of a video program, we may need to label sets of pixels or groups of image and audio frames jointly.
Machine Learning for Multimedia Content Analysis introduces machine learning techniques that are particularly powerful and effective for modeling spatial, temporal structures of multimedia data and for accomplishing common tasks of multimedia content analysis. This book systematically covers these techniques in an intuitive fashion and demonstrates their applications through case studies. This volume uses a large number of figures to illustrate and visualize complex concepts, and provides insights into the characteristics of many algorithms through examinations of their loss functions and straightforward comparisons.
Machine Learning for Multimedia Content Analysis is designed for an academic and professional audience. Researchers will find this book an invaluable tool for applying machine learning techniques to multimedia content analysis. This volume is also suitable for practitioners in industry.
Table of Contents
1 Introduction | p. 1 |
1.1 Basic Statistical Learning Problems | p. 2 |
1.2 Categorizations of Machine Learning Techniques | p. 4 |
1.2.1 Unsupervised vs. Supervised | p. 4 |
1.2.2 Generative Models vs. Discriminative Models | p. 4 |
1.2.3 Models for Simple Data vs. Models for Complex Data | p. 6 |
1.2.4 Model Identification vs. Model Prediction | p. 7 |
1.3 Multimedia Content Analysis | p. 8 |
Part I Unsupervised Learning | |
2 Dimension Reduction | p. 15 |
2.1 Objectives | p. 15 |
2.2 Singular Value Decomposition | p. 16 |
2.3 Independent Component Analysis | p. 20 |
2.3.1 Preprocessing | p. 23 |
2.3.2 Why Gaussian is Forbidden | p. 24 |
2.4 Dimension Reduction by Locally Linear Embedding | p. 26 |
2.5 Case Study | p. 30 |
Problems | p. 34 |
3 Data Clustering Techniques | p. 37 |
3.1 Introduction | p. 37 |
3.2 Spectral Clustering | p. 39 |
3.2.1 Problem Formulation and Criterion Functions | p. 39 |
3.2.2 Solution Computation | p. 42 |
3.2.3 Example | p. 46 |
3.2.4 Discussions | p. 50 |
3.3 Data Clustering by Non-Negative Matrix Factorization | p. 51 |
3.3.1 Single Linear NMF Model | p. 52 |
3.3.2 Bilinear NMF Model | p. 55 |
3.4 Spectral vs. NMF | p. 59 |
3.5 Case Study: Document Clustering Using Spectral and NMF Clustering Techniques | p. 61 |
3.5.1 Document Clustering Basics | p. 62 |
3.5.2 Document Corpora | p. 64 |
3.5.3 Evaluation Metrics | p. 64 |
3.5.4 Performance Evaluations and Comparisons | p. 65 |
Part II Generative Graphical Models | |
4 Introduction of Graphical Models | p. 73 |
4.1 Directed Graphical Model | p. 74 |
4.2 Undirected Graphical Model | p. 77 |
4.3 Generative vs. Discriminative | p. 79 |
4.4 Content of Part II | p. 80 |
5 Markov Chains and Monte Carlo Simulation | p. 81 |
5.1 Discrete-Time Markov Chain | p. 81 |
5.2 Canonical Representation | p. 84 |
5.3 Definitions and Terminologies | p. 88 |
5.4 Stationary Distribution | p. 91 |
5.5 Long Run Behavior and Convergence Rate | p. 94 |
5.6 Markov Chain Monte Carlo Simulation | p. 100 |
5.6.1 Objectives and Applications | p. 100 |
5.6.2 Rejection Sampling | p. 101 |
5.6.3 Markov Chain Monte Carlo | p. 104 |
5.6.4 Rejection Sampling vs. MCMC | p. 110 |
Problems | p. 112 |
6 Markov Random Fields and Gibbs Sampling | p. 115 |
6.1 Markov Random Fields | p. 115 |
6.2 Gibbs Distributions | p. 117 |
6.3 Gibbs - Markov Equivalence | p. 120 |
6.4 Gibbs Sampling | p. 123 |
6.5 Simulated Annealing | p. 126 |
6.6 Case Study: Video Foreground Object Segmentation by MRF | p. 133 |
6.6.1 Objective | p. 134 |
6.6.2 Related Works | p. 134 |
6.6.3 Method Outline | p. 135 |
6.6.4 Overview of Sparse Motion Layer Computation | p. 136 |
6.6.5 Dense Motion Layer Computation Using MRF | p. 138 |
6.6.6 Bayesian Inference | p. 140 |
6.6.7 Solution Computation by Gibbs Sampling | p. 141 |
6.6.8 Experimental Results | p. 143 |
Problems | p. 146 |
7 Hidden Markov Models | p. 149 |
7.1 Markov Chains vs. Hidden Markov Models | p. 149 |
7.2 Three Basic Problems for HMMs | p. 153 |
7.3 Solution to Likelihood Computation | p. 154 |
7.4 Solution to Finding Likeliest State Sequence | p. 158 |
7.5 Solution to HMM Training | p. 160 |
7.6 Expectation-Maximization Algorithm and its Variances | p. 162 |
7.6.1 Expectation-Maximization Algorithm | p. 162 |
7.6.2 Baum-Welch Algorithm | p. 164 |
7.7 Case Study: Baseball Highlight Detection Using HMMs | p. 167 |
7.7.1 Objective | p. 167 |
7.7.2 Overview | p. 167 |
7.7.3 Camera Shot Classification | p. 169 |
7.7.4 Feature Extraction | p. 172 |
7.7.5 Highlight Detection | p. 173 |
7.7.6 Experimental Evaluation | p. 174 |
Problems | p. 175 |
8 Inference and Learning for General Graphical Models | p. 179 |
8.1 Introduction | p. 179 |
8.2 Sum-product algorithm | p. 182 |
8.3 Max-product algorithm | p. 188 |
8.4 Approximate inference | p. 189 |
8.5 Learning | p. 191 |
Problems | p. 196 |
Part III Discriminative Graphical Models | |
9 Maximum Entropy Model and Conditional Random Field | p. 201 |
9.1 Overview of Maximum Entropy Model | p. 202 |
9.2 Maximum Entropy Framework | p. 204 |
9.2.1 Feature Function | p. 204 |
9.2.2 Maximum Entropy Model Construction | p. 205 |
9.2.3 Parameter Computation | p. 208 |
9.3 Comparison to Generative Models | p. 210 |
9.4 Relation to Conditional Random Field | p. 213 |
9.5 Feature Selection | p. 215 |
9.6 Case Study: Baseball Highlight Detection Using Maximum Entropy Model | p. 217 |
9.6.1 System Overview | p. 218 |
9.6.2 Highlight Detection Based on Maximum Entropy Model | p. 220 |
9.6.3 Multimedia Feature Extraction | p. 222 |
9.6.4 Multimedia Feature Vector Construction | p. 226 |
9.6.5 Experiments | p. 227 |
Problems | p. 232 |
10 Max-Margin Classifications | p. 235 |
10.1 Support Vector Machines (SVMs) | p. 236 |
10.1.1 Loss Function and Risk | p. 237 |
10.1.2 Structural Risk Minimization | p. 237 |
10.1.3 Support Vector Machines | p. 239 |
10.1.4 Theoretical Justification | p. 243 |
10.1.5 SVM Dual | p. 244 |
10.1.6 Kernel Trick | p. 245 |
10.1.7 SVM Training | p. 248 |
10.1.8 Further Discussions | p. 255 |
10.2 Maximum Margin Markov Networks | p. 257 |
10.2.1 Primal and Dual Problems | p. 257 |
10.2.2 Factorizing Dual Problem | p. 259 |
10.2.3 General Graphs and Learning Algorithm | p. 262 |
10.2.4 Max-Margin Networks vs. Other Graphical Models | p. 262 |
Problems | p. 264 |
A Appendix | p. 267 |
References | p. 269 |
Index | p. 275 |