search

business-science / gammodels

The parsnip backend for GAM Models.
https://business-science.github.io/gammodels/
Other
7 stars 2 forks source link

feature selection tuning parameter(s) #9

Open topepo opened 3 years ago

topepo commented 3 years ago

It would make sense to have a main argument for the select argument:

select: If this is TRUE then gam can add an extra penalty to each term so that it can be penalized to zero. This means that the smoothing parameter estimation that is part of fitting can completely remove terms from the model. If the corresponding smoothing parameter is estimated as zero then the extra penalty has no effect. Use gamma to increase level of penalization.

and

gamma: Increase this beyond 1 to produce smoother models. gamma multiplies the effective degrees of freedom in the GCV or UBRE/AIC. coden/gamma can be viewed as an effective sample size in the GCV score, and this also enables it to be used with REML/ML. Ignored with P-RE/ML or the efs optimizer.

Maybe use select_features for select and adjust_deg_free for gamma?

AlbertoAlmuinha commented 3 years ago

Thanks for this @topepo . This is a good idea in my opinion. Also agree with the proposed names. For my part, let's do it.

mdancho84 commented 3 years ago

I agree too! I actually wasn't aware that we can apply smooths to all of features. I do have a couple questions about factors since I'm wondering how the s() "by" argument works when automating but this is some research on my part.