Similar to the sample code for SDCA regression, we need samples for the binary classifiers:
SDCA
AveragedPerceptron
FastTree
LightGBM
There are two things to do, to resolve this issue:
I'd like to work on this one, but wouldn't want to enter in a collision with someone. Is it possible to be assigned to this one?
Thanks @bojanmisic. It doesn't look like we can assign it, but claiming it on the comments might be good enough. Tag me if you have questions.
Maybe add a new label "assigned"? @shauheen, any comment?
Well, once someone claims it, we can simply remove the help wanted and up-for-grabs tags.
@sfilipi, thanks! I believe the link to the cookbook examples has shifted, since the file has been changed recently. Would you like me to use static or dynamic API?
Dynamic is fine; I was going to log another bug for the static examples, after arranging the structure of that project.
Thanks for picking this up.
@bojanmisic checking: are you still planning to work on this?
@sfilipi, yes! I plan to work on that during the weekend. I have a strong will to contribute, just been crushed with work last two weeks. Sorry for the inconvenience, if you think someone can get it earlier, feel free to reassign.
@bojanmisic just checking, before putting the Assigned label and removing the other labels. It is yours, take your time. Take a look at PR #136. I am adding a dateset for binary trainers there.Feel free to adopt if it doesn't get merged soon.
Added example for SDCA. @sfilipi, can you take a quick look at #1456 if that is what we want - I'll proceed adding other classifiers.
Can I help on the documentation?
@bojanmisic / @rekhakodali could you please help out with some documentation and samples?
I have a table of BinaryClassification trainers that need work in #2522. You can claim StochasticDualCoordinateAscent or SymbolicStochasticGradientDescent and add yourself (removed me) as the owner on the table. You can replicate #2688 for those trainers.
@shmoradims, I'll be glad to contribute. Feel free to add me to the table for 1 or more (you can add me to both above) trainers, and I'll start working on those asap. I don't have edit permissions for the issue.
@shmoradims @bojanmisic I'll be glad to take a couple of those, as well, if it helps out.
Closing, as most samples for binary classifiers are complete, and issue #2522 captures the broader work of trainer samples.