Closed Sandy4321 closed 4 years ago
Group lasso is the regulariser you use if you know what the groups are and want to impose that information on the model. If you don't know the groups, then I guess you could cluster the groups first.
As to your reference, that is for group lasso with overlapping group, which may be useful if your covariates are not in disjoint groups. However, you should still know what the groups are to use group lasso.
I'm closing this issue as it doesn't contain any bug reports or feature request.
you don't know the groups, then I guess you could cluster the groups first. I see Do you know some code examples for this kind of case to use clusters
No I do not know of any code examples for that, but it should be quite trivial to do it. You can for example google feature clustering or biclustering.
may you help understand what to do if I do not know how many groups there in data? can your code find these groups by itself? what would be for overlapping groups like your refrences https://papers.nips.cc/paper/4275-efficient-methods-for-overlapping-group-lasso.pdf https://www.researchgate.net/publication/46582615_Fast_Overlapping_Group_Lasso