feature request for tune: Random Search over grids

[lorenzoFabbri]

Is your feature request related to a problem? Please describe.
It would be nice to be possible to perform a random search rather than an exhaustive one over the hyper-parameters' grids, something like RandomizedSearchCV in sk-learn. This is important in case the grids are fairly large.

Describe the solution you'd like

tune <- tune.spls(X, Y, ncomp = 3, 
                  test.keepX = list.keepX, test.keepY = list.keepY, random.search = TRUE)

[aljabadi]

Hi @lorenzoFabbri,

Thanks for pointing this out. This has in fact been on our list for pls-derived methods (including the block analyses). We'll keep you updated once it's in place.

[aljabadi]

Hi @lorenzoFabbri,

As an interim solution, you can tune the number of variables for X using all Y variables first, and then proceed to tune the number of variables on Y using the tuned number of variables on X. This is probably a more sensible 'fast' variation of the tune.spls which we'll implement.