matlab - libsvm with precomputed kernel: How do I compute the classification scores? -

I am working with libsvm in MATLAB and a 1-vs all SVM in the precompiled non-linear kernels. I'm a bit new to SVM and I'm trying to calculate the function function. I know that for a linear SVM, we can get (according to libsvm documentation) accordingly:

  w = model sv_coef '* model.SVs;   
 We can then calculate decision values:  
  w '* x   
 and later , Can be guessed, labeling, where  is  there is some limit; signature (w '* x + b) .  
 I am particularly interested in getting the classification scores from my non-linear kernel. How can I go about this?   
 
 
 I'm trying to understand what you mean by the classification score Compare the possibility, and choose the largest one, as I did in you  
 If you are using function function, then it is also fine. Suppose you are using RBF kernel, and  model.Label (1) = 1 , then you have (if  model.Label (1) = -1  , Then  w = -w; b = -b; )  
  [m, n] = size (model.SVs); % M is the number of support vectors, ... and the number of n features is w = model sv_coef; % M * 1 weight vector B = -model.ro; % Scalar   
 You are now given under the  v  test. And you have  [1, n] = size (v);  Then in the  i  support vector for each valve, calculate Euclidean distance (you can deflect the code given below):   i = 1  (i) = ideal for model (ssv (i, :) -v); T (I) = XP (-Gam * D (I). ^ 2); % RBF model, T1 * m vector end   
 and decision function (or score from function function):  
  s = t * W + b ;   
 You can get function functions along with other non-linear kernels.   Edit   
 With the written precomponed kernel, for example, the RBF kernel is taken:  
 < Code>% RBF kernel: XP (-Gam * * UV | ^ 2) RBF = @ (X, Y) XP (-Gm. * Pdist2 (X, Y, 'Euclidean'). ^ 2); The sample with serial matrix is ​​required as the first column in the form of serial number K_train = [(1: numTrain) ', RBF (train data, train data)]; K_test = [(1: numTest) ', RBF (TestData, Train Data)]; % # Train and test model = svmtrain (train labels, ktrain, '-4 4); [Pradelabel, ~, ~] = svmpredict (testlabel, k_test, model); % # Illusion matrix C = confusion (test label, predLabel);