Which of the following classifiers do not require model selection?

Nearest neighbor classifiers.

A linear maximum margin classifiers.

Decision tree classifiers.

None of the above require model selection.

All of the above require model selection.

Cross validation, applied to a given classifier, can be used for various purposes, but it always estimates a specific quantity. Which one?

The classifier parameters.

The optimal complexity of the model.

The prediction error of the classifier.

The model hyperparameters.

Recall that K-fold cross validation removes a validation set, then splits up the remaining data into K (roughly) equally sized blocks. How large should the blocks be chosen?

Each block should be as large as possible.

Each block should be as small as possible.

There is a trade-off, since both large and small blocks have advantages.

The block size does not matter, as long as the split is random.