A Support Vector Machines

Embed Size (px)

Citation preview

  • 7/30/2019 A Support Vector Machines

    1/2

    A Support vector machines (SVMs) are a family of supervised machine learning methods

    developed in recent years. SVMs were first suggested by Vapnik in the 1960s for classification.

    The main idea behind the SVM algorithm is to map input vectors into a feature space of higher

    dimension, construct a linear decision surface (hyperplane), and then optimize that hyperplane

    for generalization. SVMs are used for classification, regression, and ranking, and are used for

    related tasks such information retrieval and optical character recognition.The support vector machine (SVM)is a training algorithm for learning classification and

    regression rules from data. SVMs arose from statistical learning theory; the aim being to solve

    only the problem of interest without solving a more difficult problem as an intermediate step.

    SVMs are based on the structural risk minimization principle, closely related to regularization

    theory.

    Two key elements in the implementation of SVM are the techniques of mathematical

    programming and kernel functions. The parameters are found by solving a quadratic

    programming problem with linear equality and inequality constraints; rather than by solving a

    non-convex, unconstrained optimization problem. The kernel forms the basic parts of SVM and

    the flexibility of kernel functions allows the SVM to search a wide variety of hypothesis spaces.

    The basic principles of an SVM are as follows1. The SVM performs a nonlinear mapping of the input vectors (objects) from the input space Rdinto a high dimensional feature space H; the mapping is determined by a kernel function.

    2. It finds a linear decision rule in the feature space H in the form of an optimal separating plane,

    which is the one that leaves the widest margin between the decision boundary and the input

    vectors mapped into H.

    3. This plane is found by solving the following constrained quadratic programming problem:

    Maximize W

    1 1 1

    1,

    2

    n n n

    i i j i j i j

    i i j

    y y K x x

    under the constraints1

    0n

    i i

    i

    y

    and

    0 C for i=1,2,.., n wherei dx R are the training sample set (TSS) vectors, and

    1, 1iy the corresponding class labels. C is a constant needed for non-separable classes;

    ,K u v is a kernel function.There are three types of support vector machines:

    Polynomial SVM:

    , , 1d

    K u v u v

    This type of machine describes the mapping of an input space into the feature space that

    contains all products .....i i j j k k u v u v u v up to degree d.

    Radial Basis Function SVM:

    2, expK u v u v

    This type of machine forms a separation rule according to radial hyperplanes.

    Two-layer neural network SVM:

    , tanh ,K u v w u v c This type of machine generates a kind of two-layer neural network without back propagation.

  • 7/30/2019 A Support Vector Machines

    2/2