Logloss SGD

121 views
Skip to first unread message

martij...@gmail.com

unread,
Oct 4, 2016, 7:26:20 AM10/4/16
to libFM - Factorization Machines
Hi there, I'm trying to create a logloss classification benchmark using SGD, though as far as I've seen there is only an accuracy measure using sgd. Is this correct and/or anyone has a solution for this problem?

Best regards, Martijn

Alexey Renyov

unread,
Oct 9, 2016, 4:26:47 PM10/9/16
to libFM - Factorization Machines, martij...@gmail.com
Yep.

But you can extract your predictions and calculate any metric you want using any tool like python or R.
Other variant is to modify evaluate function here https://github.com/srendle/libfm/blob/master/src/libfm/src/fm_learn.h#L104 .
So for example you can calculate here log loss if you need or any other metric.
Reply all
Reply to author
Forward
0 new messages