apply it to regression - but I agree that pretty much reduces to what you did!

Thanks to the pointer on the log det and trace approxes. I'll definitely have a look! (2/2)

Yeah pretty much! I'm not entirely sure how I see how it's an approximation of H[Y_1, ..., Y_B | X_1, ..., X_B] - could you clarify?

I agree it doesn't take into account aleatoric uncertainty - but that's kind of the point imo! My focus here was really on classification, although you can ... (1/2)

I'll also be presenting "Active learning for affinity prediction of antibodies" at both BDU and MLSB (tomorrow), work done with Alexandra Gessner, Owen Vickery, Dino Oglic, and Talip Ucar! openreview.net/pdf?id=rSnUP... (3/3)

This project was a lot of fun, but wouldn't have been possible without my great collaborators: @spmontecarlo.bsky.social, @tomdiethe.com, and Henry Moss.

Come to East Rooms 8/15 for the lightning talks (14:30-15:00) and poster session (15:00-16:00), but also for a great workshop! (2/3)

Could you add me please? Doing ML for antibody therapeutics!

✋