Course No.：S070105ZY001
Course Category：Professional Course
Period/Credits：40/2
Prerequisites：Higher Mathematics.
Aims & Requirements：
This course focuses on support vector machines (SVM), which is a general method for data mining and machine learning. Support vector machines can deal with pattern recognition (classification, discriminant analysis), regression problems (time series analysis) and many other issues very successfully. It can also promote the field of prediction and evaluation. The basic idea of support vector machine is to transform learning problem into optimization problem. This course will systematically introduce the theoretical basis of support vector machines (including the optimization of the basic theory, the focus in the introduction of convex programming and statistical learning theory of the theoretical foundation and nuclear), introduce the basic ideas of how to use support vector machine to solve classification problems, regression problems, semi-supervised problem , unsupervised problem and the corresponding algorithms, introduces support vector machine’s expansion model and its application in practical problems.
In this course, students are hoped to get a more comprehensive understanding of the support vector machine, to initially master the usage and basic skills of using support vector machine to solve practical problems.
Primary Coverage：
Section 1: Optimization Basis
Optimization problems on Euclidean space and Hilbert space; Convex programming on Euclidean space and Hilbert space; Convex programming with generalized inequality constraints on the European space and Hilbert space (cone programming, semi-definite programming, etc.).
Section 2: Linear Classifier
Classification problem; SVM of the linearly separable problem; linear support vector machine;
Section 3: Linear Regression
Linear regression problem; hard e-strip ultra-flat; hard e-strip support vector regression machine; linear e-strip support vector regression machine learning;
Section 4: Nuclear and support vector machine
Linear programming and nonlinear programming; kernel function; support vector machine and its nature; the selection of kernel function in support vector machines;
Section 5: Statistical learning theory of C-support vector machine classifier
Statistical formulation of the classification; empirical risk minimization principle; the VC dimension; structural risk minimization principle; statistical learning theory of the C-support vector machine classifier.
Author：Yingjie Tian (CAS Research Center On Fictitious Economy & Data Science)
Date：June, 2009
