Regarding the training dataset size while fitting a svm classifier about secml HOT 2 CLOSED

Hi @m-melis, thank you for your quick reply. I'm clear now. Today is the first day of using this library. I shall definitely report i find anything different. Thank you again.

from secml.

Hi @fugokidi, thanks for raising this issue.

The CClassifierSVM.fit(x, y) method should fit the model on the whole dataset. Instead, while using the CClassifierSVM.estimate_parameters() method to find the best parameters you can customize the number of folds in the num_folds parameter of the CDataSplitter you are passing.

There is also an additional method of CClassifierSVM, called fit_forward which currently uses a CDataSplitterKFold with a fixed number of folds equal to 5 and cannot be changed.

This should be the expected behavior. If you find anything different please let us know.

from secml.