Currently, the results are on the level of a cv fold. We are storing indices etc. to be able to plot each fold for each model. However, RandomizedSearchCV and BayesSearchCV don't actually return that level of detail, just means across folds. Therefore, we run cross_validate after parameter search. This leads to unnecessary compute, as the relevant models have been cross-validated already. We could either decide to not store fold-wise data (just means) or to implement search manually using cross_validate. The former is probably ok.
Currently, the results are on the level of a cv fold. We are storing indices etc. to be able to plot each fold for each model. However,
RandomizedSearchCVandBayesSearchCVdon't actually return that level of detail, just means across folds. Therefore, we runcross_validateafter parameter search. This leads to unnecessary compute, as the relevant models have been cross-validated already. We could either decide to not store fold-wise data (just means) or to implement search manually usingcross_validate. The former is probably ok.