MatLab自带的SVM，快速入门使用

MatLab自带的SVM一共集成了两个函数： svmtrain和 svmclassify。

1.首先查看你的MatLab版本是否已经集成了SVM，我的版本R2013a，主要参考MatLab帮助文档(自带demo)

>> help svmtrain

 svmtrain Train a support vector machine classifier

    SVMSTRUCT = svmtrain(TRAINING, Y) trains a support vector machine (SVM)

    classifier on data taken from two groups. TRAINING is a numeric matrix

    of predictor data. Rows of TRAINING correspond to observations; columns

    correspond to features. Y is a column vector that contains the known

    class labels for TRAINING. Y is a grouping variable, i.e., it can be a

    categorical, numeric, or logical vector; a cell vector of strings; or a

    character matrix with each row representing a class label (see help for

    groupingvariable). Each element of Y specifies the group the

    corresponding row of TRAINING belongs to. TRAINING and Y must have the

    same number of rows. SVMSTRUCT contains information about the trained

    classifier, including the support vectors, that is used by SVMCLASSIFY

    for classification. svmtrain treats NaNs, empty strings or 'undefined'

    values as missing values and ignores the corresponding rows in

    TRAINING and Y.

2.下面是使用文档：

 SVMSTRUCT = svmtrain(TRAINING, Y, 'PARAM1',val1, 'PARAM2',val2, ...)

    specifies one or more of the following name/value pairs:

 

       Name                Value

       'kernel_function'  A string or a function handle specifying the

                          kernel function used to represent the dot

                          product in a new space. The value can be one of

                          the following:

                          'linear'     - Linear kernel or dot product

                                         (default). In this case, svmtrain

                                         finds the optimal separating plane

                                         in the original space.

                          'quadratic'  - Quadratic kernel

                          'polynomial' - Polynomial kernel with default

                                         order 3. To specify another order,

                                         use the 'polyorder' argument.

                          'rbf'        - Gaussian Radial Basis Function

                                         with default scaling factor 1. To

                                         specify another scaling factor,

                                         use the 'rbf_sigma' argument.

                          'mlp'        - Multilayer Perceptron kernel (MLP)

                                         with default weight 1 and default

                                         bias -1. To specify another weight

                                         or bias, use the 'mlp_params'

                                         argument.

                          function     - A kernel function specified using

                                         @(for example @KFUN), or an

                                         anonymous function. A kernel

                                         function must be of the form

 

                                         function K = KFUN(U, V)

 

                                         The returned value, K, is a matrix

                                         of size M-by-N, where M and N are

                                         the number of rows in U and V

                                         respectively.

 

    'rbf_sigma'           A positive number specifying the scaling factor

                          in the Gaussian radial basis function kernel.

                          Default is 1.

 

    'polyorder'           A positive integer specifying the order of the

                          polynomial kernel. Default is 3.

 

    'mlp_params'          A vector [P1 P2] specifying the parameters of MLP

                          kernel.  The MLP kernel takes the form:

                          K = tanh(P1*U*V' + P2),

                          where P1 > 0 and P2 < 0. Default is [1,-1].

 

    'method'              A string specifying the method used to find the

                          separating hyperplane. Choices are:

                          'SMO' - Sequential Minimal Optimization (SMO)

                                  method (default). It implements the L1

                                  soft-margin SVM classifier.

                          'QP'  - Quadratic programming (requires an

                                  Optimization Toolbox license). It

                                  implements the L2 soft-margin SVM

                                  classifier. Method 'QP' doesn't scale

                                  well for TRAINING with large number of

                                  observations.

                          'LS'  - Least-squares method. It implements the

                                  L2 soft-margin SVM classifier.

 

    'options'             Options structure created using either STATSET or

                          OPTIMSET.

                          * When you set 'method' to 'SMO' (default),

                            create the options structure using STATSET.

                            Applicable options:

                            'Display'  Level of display output.  Choices

                                     are 'off' (the default), 'iter', and

                                     'final'. Value 'iter' reports every

                                     500 iterations.

                            'MaxIter'  A positive integer specifying the

                                     maximum number of iterations allowed.

                                     Default is 15000 for method 'SMO'.

                          * When you set method to 'QP', create the

                            options structure using OPTIMSET. For details

                            of applicable options choices, see QUADPROG

                            options. SVM uses a convex quadratic program,

                            so you can choose the 'interior-point-convex'

                            algorithm in QUADPROG.

 

   'tolkkt'              A positive scalar that specifies the tolerance

                         with which the Karush-Kuhn-Tucker (KKT) conditions

                         are checked for method 'SMO'. Default is

                         1.0000e-003.

 

   'kktviolationlevel'   A scalar specifying the fraction of observations

                         that are allowed to violate the KKT conditions for

                         method 'SMO'. Setting this value to be positive

                         helps the algorithm to converge faster if it is

                         fluctuating near a good solution. Default is 0.

 

   'kernelcachelimit'    A positive scalar S specifying the size of the

                         kernel matrix cache for method 'SMO'. The

                         algorithm keeps a matrix with up to S * S

                         double-precision numbers in memory. Default is

                         5000. When the number of points in TRAINING

                         exceeds S, the SMO method slows down. It's

                         recommended to set S as large as your system

                         permits.

 

   'boxconstraint'       The box constraint C for the soft margin. C can be

                         a positive numeric scalar or a vector of positive

                         numbers with the number of elements equal to the

                         number of rows in TRAINING.

                         Default is 1.

                         * If C is a scalar, it is automatically rescaled

                           by N/(2*N1) for the observations of group one,

                           and by N/(2*N2) for the observations of group

                           two, where N1 is the number of observations in

                           group one, N2 is the number of observations in

                           group two. The rescaling is done to take into

                           account unbalanced groups, i.e., when N1 and N2

                           are different.

                         * If C is a vector, then each element of C

                           specifies the box constraint for the

                           corresponding observation.

 

    'autoscale'          A logical value specifying whether or not to

                         shift and scale the data points before training.

                         When the value is true, the columns of TRAINING

                         are shifted and scaled to have zero mean unit

                         variance. Default is true.

 

    'showplot'           A logical value specifying whether or not to show

                         a plot. When the value is true, svmtrain creates a

                         plot of the grouped data and the separating line

                         for the classifier, when using data with 2

                         features (columns). Default is false.

 

    SVMSTRUCT is a structure having the following properties:

 

    SupportVectors       Matrix of data points with each row corresponding

                         to a support vector. 

                         Note: when 'autoscale' is false, this field

                         contains original support vectors in TRAINING.

                         When 'autoscale' is true, this field contains

                         shifted and scaled vectors from TRAINING.

    Alpha                Vector of Lagrange multipliers for the support

                         vectors. The sign is positive for support vectors

                         belonging to the first group and negative for

                         support vectors belonging to the second group.

    Bias                 Intercept of the hyperplane that separates

                         the two groups.

                         Note: when 'autoscale' is false, this field

                         corresponds to the original data points in

                         TRAINING. When 'autoscale' is true, this field

                         corresponds to shifted and scaled data points.

    KernelFunction       The function handle of kernel function used.

    KernelFunctionArgs   Cell array containing the additional arguments

                         for the kernel function.

    GroupNames           A column vector that contains the known

                         class labels for TRAINING. Y is a grouping

                         variable (see help for groupingvariable).

    SupportVectorIndices A column vector indicating the indices of support

                         vectors.

    ScaleData            This field contains information about auto-scale.

                         When 'autoscale' is false, it is empty. When

                         'autoscale' is set to true, it is a structure

                         containing two fields:

                         shift       - A row vector containing the negative

                                       of the mean across all observations

                                       in TRAINING.

                         scaleFactor - A row vector whose value is

                                       1./STD(TRAINING).

    FigureHandles        A vector of figure handles created by svmtrain

                         when 'showplot' argument is TRUE.

 

    Example:

        % Load the data and select features for classification

        load fisheriris

        X = [meas(:,1), meas(:,2)];

        % Extract the Setosa class

        Y = nominal(ismember(species,'setosa'));

        % Randomly partitions observations into a training set and a test

        % set using stratified holdout

        P = cvpartition(Y,'Holdout',0.20);

        % Use a linear support vector machine classifier

        svmStruct = svmtrain(X(P.training,:),Y(P.training),'showplot',true);

        C = svmclassify(svmStruct,X(P.test,:),'showplot',true);

        errRate = sum(Y(P.test)~= C)/P.TestSize  %mis-classification rate

        conMat = confusionmat(Y(P.test),C) % the confusion matrix

 

    See also svmclassify, NaiveBayes, classregtree, classify, TreeBagger,

             groupingvariable

    Reference page in Help browser

       doc svmtrain

文档最佳观看效果：在MatLab中使用doc svmtrain命令

3.一个例子：Examples

Find a line separating the Fisher iris data on versicolor and

virginica species, according to the petal length and petal width measurements.

These two species are in rows 51 and higher of the data set, and the

petal length and width are the third and fourth columns.

load fisheriris

xdata = meas(51:end,3:4);

group = species(51:end);

svmStruct = svmtrain(xdata,group,'showplot',true);

也可以在线查看文档：http://www.mathworks.com/help/releases/R2013a/stats/svmtrain.html

另外，这个写的不错(例子写的非常好)：

《(Matlab)SVM工具箱快速入手简易教程》

ps.关于例子，百度文库里有结果图可看，其他各种转载的博客上无图。

还有一个博客介绍（我没试，不知道文中要下载的svm和MatLab自带的svm是不是一回事）：

http://zyy554221.blog.sohu.com/82115143.html?qq-pf-to=pcqq.c2c