TABLE
OF CONTENTS
1
Introduction
2
Support Vector Machines in Classification and Regression  An
Introduction
3
Iterative Single Data Algorithm for Kernel Machines from Huge
Data Sets: Theory and Performace
4
Feature Reduction with Support Vector Machines and Application
in DNA Microarray Analysis
5
Semisupervised Learning and Applications
6
Unsupervised Learning by Principal and Independent Component
Analysis
Appendices
A
Support Vector Machines
B
Matlab Code for ISDA Classification
D
Matlab Code for Conjugate Gradient Method with Box Constraints
E
Uncorrelatedness and Independence
F
Independent Component Analysis by Empirical Estimation of Score
Functions i.e., Probability Density Functions
G
SemiL User Guide
1
Introduction
1.1
An Overview of Machine Learning
1.2
Challenges in Machine Learning
1.2.1
Solving LargeScale SVMs
1.2.2
Feature Reduction with Support Vector Machines
1.2.3
GraphBased Semisupervised Learning Algorithms
1.2.4
Unsupervised Learning Based on Principle of Redundancy
Reduction
2
Support Vector Machines in Classification and Regression  An
Introduction
2.1
Basics of Learning from Data
2.2
Support Vector Machines in Classification and Regression
2.2.1
Linear Maximal Margin Classifier for Linearly Separable Data
2.2.2
Linear Soft Margin Classifier for Overlapping Classes
2.2.3
The Nonlinear SVMs Classifier
2.2.4
Regression by Support Vector Machines
2.3
Implementation Issues
back
to top
3
Iterative Single Data Algorithm for Kernel Machines from Huge
Data Sets: Theory and Performance
3.1
Introduction
3.2
Iterative Single Data Algorithm for Positive Definite Kernels
without Bias Term b
3.2.1
Kernel AdaTron in Classification
3.2.2
SMO without Bias Term b in Classification
3.2.3
Kernel AdaTron in Regression
3.2.4
SMO without Bias Term b in Regression
3.2.5
The Coordinate Ascent Based Learning for Nonlinear
Classification and Regression Tasks
3.2.6
Discussion on ISDA Without a Bias Term b
3.3
Iterative Single Data Algorithm with an Explicit Bias Term b
3.3.1
Iterative Single Data Algorithm for SVMs Classification with a
Bias Term b
3.4
Performance of the Iterative Single Data Algorithm and
Comparisons
3.5
Implementation Issues
3.5.1
Workingset Selection and Shrinking of ISDA for Classification
3.5.2
Computation of the Kernel Matrix and Caching of ISDA for
Classification
3.5.3
Implementation Details of ISDA for Regression
3.6
Conclusions
4
Feature Reduction with Support Vector Machines and Application
in DNA Microarray Analysis
4.1
Introduction
4.2
Basics of Microarray Technology
4.3
Some Prior Work
4.3.1
Recursive Feature Elimination with Support Vector Machines
4.3.2
Selection Bias and How to Avoid It
4.4
Influence of the Penalty Parameter C in RFESVMs
4.5
Gene Selection for the Colon Cancer and the Lymphoma Data Sets
4.5.1
Results for Various C Parameters
4.5.2
Simulation Results with Different Preprocessing Procedures
4.6
Comparison between RFESVMs and the Nearest Shrunken Centroid
Method
4.6.1
Basic Concept of Nearest Shrunken Centroid Method
4.6.2
Results on the Colon Cancer Data Set and the Lymphoma Data Set
4.7
Comparison of Genes’ Ranking with Different Algorithms
4.8
Conclusions
back
to top
5
Semisupervised Learning and Applications
5.1
Introduction
5.2
Gaussian Random Fields Model and Consistency Method
5.2.1
Gaussian Random Fields Model
5.2.2
Global Consistency Model
5.2.3
Random Walks on Graph
5.3
An Investigation of the Effect of Unbalanced labeled Data on CM
and GRFM Algorithms
5.3.1
Background and Test Settings
5.3.2
Results on the Rec Data Set
5.3.3
Possible Theoretical Explanations on the Effect of Unbalanced
Labeled Data
5.4
Classifier Output Normalization: A Novel Decision Rule for
Semisupervised Learning Algorithm
5.5
Performance Comparison of Semisupervised Learning Algorithms
5.5.1
Low Density Separation: Integration of GraphBased Distances
and rTSVM
5.5.2
Combining GraphBased Distance with Manifold Approaches
5.5.3
Test Data Sets
5.5.4
Performance Comparison Between the LDS and the Manifold
Approaches
5.5.5
Normalizatioin Steps and the Effect of σ
5.6
Implementation of the Manifold Approaches
5.6.1
Variants of the Manifold Approaches Implemented in the
Software Package SemiL
5.6.2
Implementation Details of SemiL
5.6.3
Conjugate Gradient Method with Box Constraints
5.6.4
Simulation Results on the MNIST Data Set
5.7
An Overview of Text Classification
5.8
Conclusions
6
Unsupervised Learning by Principal and Independent Component
Analysis
6.1
Principal Component Analysis
6.2
Independent Component Analysis
6.3
Concluding Remarks
A
Support Vector Machines
A.1
L2 Soft Margin Classifier
A.2
L2 Soft Regressor
A.3
Geometry and the Margin
B
Matlab Code for ISDA Classification
C
Matlab Code for ISDA Regression
D
Matlab Code for Conjugate Gradient Method with Box Constraints
E
Uncorrelatedness and Independence
F
Independent Component Analysis by Empirical Estimation of Score
Functions i.e., Probability Density Functions
G
SemiL User Guide
G.1
Installation
G.2
Input Data Format
G.2.1
Raw Data Format
G.3
Getting Started
G.3.1
Design Stage
Index
References
back
to top
