SVM Grid Search

Performing grid search for SVM, using the default Matlab toolbox.

The main function svm_grid_search, preforms a grid search using the following parameters: name of the kernel to be used, values for the kernel, values for the boxconstraint, and values for the kktviolatonlevel level. It gives as output array with the trained SVMs, array showing if a SVM was not able to be trained (converge), and the accuracy for the train, validation and test sets. The function is easy to upgrade and mode.
You can use structures for I/O (in svm_grid_search_structure), or the standard way with listing all the variables (in svm_grid_search). For me the use of structures is more elegant, when having to pass a lot of the same variables between function, and it is easier to mode.