sklearn contains distinct estimators that have built in cross-validation support. From the sklearn docs:
Cross-validation estimators are named EstimatorCV and tend to be roughly equivalent to GridSearchCV(Estimator(), ...). The advantage of using a cross-validation estimator over the canonical estimator class along with grid search is that they can take advantage of warm-starting by reusing precomputed results in the previous steps of the cross-validation process.
Some third-party libraries like nilearn use the CV versions of these estimators in their codebase. This means that today these libraries can't be easily accelerated with cuml.accel.
We should evaluate what the level of effort would be to add support for these estimator types to cuml and cuml.accel, potentially as a follow-up to #7820.
There are many CV estimators but these seem like they would be the most valuable and leverage base estimators that are already implemented in cuml:
This notebook could be useful for validation once we have implemented these estimators.
sklearn contains distinct estimators that have built in cross-validation support. From the sklearn docs:
Some third-party libraries like nilearn use the CV versions of these estimators in their codebase. This means that today these libraries can't be easily accelerated with cuml.accel.
We should evaluate what the level of effort would be to add support for these estimator types to cuml and cuml.accel, potentially as a follow-up to #7820.
There are many CV estimators but these seem like they would be the most valuable and leverage base estimators that are already implemented in cuml:
This notebook could be useful for validation once we have implemented these estimators.