Our new paper: ⚡Zapping!⚡
Rehearsing transfer during pre-training by repeatedly "zapping" (resampling) weights is a cheap form of meta-learning that creates neural nets that are better at transfer learning!
w/ @fratietal, @jeffclune, and @CheneyLab:
arxiv.org/abs/2310.07996
Congratulations to Nick Cheney @CheneyLab, who was just awarded an NSF CAREER award, one of the highest-level recognitions that a young faculty member can receive.
He proposes to make AI more accessible via AutoML, NAS, and robotics. Details here:
nsf.gov/awardsearch/showAwar…