|
I'm looking for practical multi-task (multi-response) linear regression code, that can run on dense data, with a few 1000s of samples and 500K to 2M features. Doesn't matter if it's kernel methods, penalized regression, whatever. I can throw about 40 cores and 512GB RAM at it, as long as doesn't take weeks (taking into account cross-validation and tuning of parameters). I want to get actual results from this, this is not a bakeoff to see which method converges faster on some toy problem ;) |
