|
For a project i have to design a binary classifier based on experimental data. The missclassification cost is unfortunately not symmetric. I found two matlab functions 'svmtrain' and 'fitcsvm'. Unfortunately neither of these functions combines what I need. SVMtrain allows a lot of tuning but no assymetric costs. fitsvm allows assymetric costs but i can't chose sigma in the 'rbf' kernel which i found out is cruicial. Is there a possiblity to use both? |
