Cross-validation and hyperparameter tuning

Module 3. Hyperparameter tuning Automated tuning
#1

In the Cross-validation and hyperparameter tuning notebook, why do we have to pass the cross validation parameter (cv=2) in the GridSearchCV initialization?

model_grid_search = GridSearchCV(model, param_grid=param_grid, n_jobs=4, cv=2)

In the next line we are anyway going to cross validate, so I found it redundant and complex.

cv_results = cross_validate(
    model_grid_search, data, target, cv=3, return_estimator=True)

Have I understood this correctly? If not, could someone explain why this is necessary?
Thank you

1 Like
#2

Alright, so after a few days I came across this article that explores the concept of nested cross-validation, which cleared my doubt.

2 Likes