search

QB3 / sparse-ho

Fast hyperparameter settings for non-smooth estimators:
http://qb3.github.io/sparse-ho
BSD 3-Clause "New" or "Revised" License
38 stars 15 forks source link

Comments on Weighted Lasso example with held-out test set #95

Open ksehic opened 3 years ago

ksehic commented 3 years ago

Hi @QB3 @mathurinm

About the weighted Lasso example, it seems you are using kfold for training sparse-ho crossval function, then why are you passing indices in heldout function line 80? It should be HeldOutMSE(None, None)?

It would be nice when you are splitting to have X_train y_test instead of passing again X and y. I have noticed that the number of training samples could be too small. As you can see if we have more samples, sparse-ho is doing better....

image

image

QB3 commented 3 years ago

It should be HeldOutMSE(None, None)?

Yes you are right! thx!

QB3 commented 3 years ago

Btw I think I should replace 'WeightedLasso val' by 'WeightedLassoCV' wdyt?

mathurinm commented 3 years ago

and put LassoCV test, too ?