This is an interesting new paper on the topic of feature learning. They show that several aspects of feature learning in neural nets can be reproduced in a simpler kernel-based model dubbed a “Recursive Feature Machine” (RFM) https://t.co/J6xjrRHL9i
TheoretiCS has launched its first issue! https://t.co/AEBgg4ezcN
TheoretiCS is a new, peer-reviewed Diamond Open Access electronic journal covering all areas of Theoretical Computer Science. Looking forward to seeing how the journal develops.
One of my pet peeves is this thing in US culture which urges people to come up with a "cool" story of their life, starting from a poor miserable state, shining into a successful career. It pushes people to come up with bipolar exaggerated narrative of their life ....
The Twitter war against Mastodon in general continues. Various links to Mastodon sites are now being marked as potentially unsafe.
I was going to post a link to the discussion, but Twitter won't let me. So I've charted it so you can see it:
https://t.co/MoNtPQRutU
🧵New paper w Nicholas Carlini & @florian_tramer: "Considerations for Differentially Private Learning with Large-Scale Public Pretraining."
We critique the increasingly popular use of large-scale public pretraining in private ML.
Comments welcome.
https://t.co/Gh4CmBzMNm 1/n
We're excited to welcome Lawrence Saul as our next speaker on Thursday Dec 15 at 4pm!
Title: A Geometrical Connection Between Sparse and Low-rank Matrices and Its Uses for Machine Learning
Register here: https://t.co/Lp8BQHbk4B