problem in plottin kernel svm inplementation

Hi,

I am implementing the Kernel SVM for a non-linear separable data set. I have implemented the algorithm for 4 points to see the results and then generalize for a larger data set. When I plot the results, I can see the surfaces separating the data sets but for a larger data set it does not give proper results, so I am not sure if I am plotting in the right way. Could someone please let me know how to plot the results.

Thanks in advance, Sharmi

I have added the Kernel SVM code below

% kernel function g = @(x,y) (x*y'+1).^2; X_data = cat(1,data_set(1:N1,1:2),-data_set(N1+1:N1+N2,1:2)); y = cat(1,data_set(1:N1,3),data_set(N1+1:N1+N2,3)); H = (y*y').*g(X_data,X_data); f=-ones(N1+N2,1); Aeq=y'; beq=0; lb=zeros(N1+N2,1); ub=C*ones(N1+N2,1); lambda=quadprog(H,f,[],[],Aeq,beq,lb,ub);

lambda(floor(10000*abs(lambda))==0)=0; s =g(X_data,X_data);

% weight vector w = g(data_set,data_set)*(y.*lambda); q=find(lambda>0 & lambda<C-.000001);

x_min=-1+floor(min(abs(g(data_set,data_set)))); x_max=1+ceil(max(abs(g(data_set,data_set)))); y_min=-1+floor(min(abs(data_set(:,2)))); y_max=1+ceil(max(abs(data_set(:,2))));

xi=x_min:x_max; for i=1:N1+N2 yi=(y.*lambda)*g(data_set(i,:),data_set); end

xi= xi(1:size(yi));

axis equal figure plot(xi,yi,'k') figure mesh(yi);

data_set is the 2D non-linearly separable random data set which is augmented for each class and normalized for the second class.

[Shashank Prasanna]

Could you shed more light on "does not give proper results"? Do you get an error message? What seems to be the issue?

[rini]

I want to plot my data set along with the results of Kernel svm in the same figure. However, with this code, if I try to plot the data and the hyper-planes, I get an error that vectors should be of same length. I want to know if I could show the hyper-planes separating the non-linear data set. Other than this, I am not getting any error in the code

Thanks