On Fri, 6 Dec 2019, Sven Schreiber wrote:
Am 05.12.2019 um 23:22 schrieb Allin Cottrell:
> I've been working lately on a lasso package for gretl. It's not at the
> point of uploading for approval (among other things, it doesn't have a
> GUI hook yet) but I'd be grateful if anyone is willing to try some
> testing and give me feedback.
Hi Allin, great news of course!
I haven't tested yet, just looked at the pdf and code a little. Very
interesting that you are doing a hansl package for that instead of pure
C, at least as the middle layer.
I thought that approach worked pretty well for the dbnomics addon.
In the lasso case too the middle layer (which I don't think is
time-critical) is less tedious to write in hansl, and more easily
revised.
One question about the syntax w.r.t.
the hidden backend:
err = _admm_lasso(X, Y, ret)
apparently stuffs results into 'ret'. From a Hansl point of view it is
of course strange that you can pass ret without pointerizing it then.
If it were a public function I'd use the pointerized form for @ret
to make the modifiability of the argument clear. But it's not
actually necessary since we're not creating a bundle from scratch,
just modifying its content.
Another question concerning your speed discussion in the pdf: The
numbers sound quite good. But you just say "glmnet is faster" without
any number there. What factor are we roughly talking about - 50%? 100%?
Well, it depends -- I think, mostly on the number of lambda values
used in cross validation. In my bigger tests glmnet is about twice
as fast.
Allin