01-Sep-2016 10:51 AM
Matlab provides the supervised learning functions: fitcsvm, fitctree, fitrtree, fitensemble, fitcknn, fitcnb, fitcdiscr, etc. using the name-value pair option fit_function(..., 'CrossVal', 'on', ...) one can perform k-fold cross validation (with default k = 10).
However, I cannot extract the tuning parameter(s) for several methods, like C in SVM, etc., and performance indicators, like AUC, etc.
Again, how to run all these fitting functions into a grid of optimal parameter(s) search, which would be the next obvious question? I guess Matlab is using some default tuning ....