I'm running on k8s, and I'm confused what will be better (faster):

• running grid search hyper-parameter method with gridsearchcv and n_jobs=-1 or
• using map-reduce technique (map different parameters to different nodes and select the best score (reduce)

What will be faster ?

In the first method (gridsearchcv) I can use sklearn (no need to implement)

In the second method (map-reduce) I need to implement the map-reduce alg