Why does grid search cross validation give same value of mean square error for different values of C and gamma in support vector regression ?

2 次查看（过去 30 天）

Sharda 2016-2-10

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/267450-why-does-grid-search-cross-validation-give-same-value-of-mean-square-error-for-different-values-of-c

评论： Aagya Niraula 2019-2-17

I am using libsvm in matlab for time series prediction using support vector regression . When I use grid search cross validation to select parameters C and gamma, the value of cross validation mean square error is coming same for different values of these parameters.So,by default the best C and gamma are the first values in the given range of parameters which is clearly not the case.

How can I sort this issue and find best value of parameters ?

My code is as follows,

[C,gamma] = meshgrid( -10:1:10, -10:1:10);

for j=1:numel(C) mse_cv(j) = svmtrain(svm_label,svm_data, ... sprintf('-s %d -t %d -c %f -g %f -p %f -v %d -h %d ',s,t, 2^C(j), 2^gamma(j),eps, folds,h )); end

here,

svm_label =

49.6665 49.6665 49.6668 49.6670 49.6671

and

svm_data =

49.6664 49.6665 49.6665 49.6668 49.6670

eps=0.005 ,t=2,s=3,v=5 and h=0.