Hi everyone,
We're excited to announce that the unlearning challenge is now open for submissions on Kaggle! We will accept submissions until November 27, 2023.
We recommend using this starting notebook for making a submission, and simply replacing the contents of the `unlearning` function. Please see the Kaggle page for more information on how to make a submission.
We are also excited to release our write-up with a formal definition of unlearning, linking it to Differential Privacy, and a description of the metric and evaluation protocol that we use in the challenge.
We look forward to receiving your submissions!
Thanks,
Eleni, on behalf of the unlearning challenge organizers
Hi,Thank you for launching the competition! My team is very eager to get started. I just had a quick question regarding the 512 check points. Is thatA) A way to cross validate across 512 splits? I'm not sure if the retain, forget or validation loaders are being modified so I think it might be B) It's doing 512 runs of the same experiment/split? Could you please clarify at your earliest convenience?Thanks again,Mohit
--
You received this message because you are subscribed to the Google Groups "unlearning-challenge" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unlearning-chall...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unlearning-challenge/CANQd_MMbpQOU%3Dw8Nje%2BxendJ_qwfivf0wf_rN14NULwBOJv56w%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.