I annotated and reimplemented Norm-Agnostic Residual Networks (NAG), by Tomas Figliolia & @BerenMillidge at @ZyphraAI, and reproduced its core geometry: controlled norm growth, steadier rotation with depth, and fewer attention sinks.
Blog: alexanderspeicher.com/blog/norm-agno…
@JohnSmith23_4@PentelEnergel Chapter 3 was definitely the toughest of the first 4. I did a lot of reviewing/quizzing with claude before tackling the questions but even then it took me a long time
Hi all, I'm new here but I'm starting the year off replicating ML papers on a 2-3 week cadence
I started off with Alexnet since it represented large a shift in machine learning!
github.com/wtoth/alexnet
p.s. I am also working my way through Understanding Deep Learning
#ml
6K Followers 305 FollowingCo-founder and CEO at Elorian AI.
Ex (director) at Google DeepMind after 12y. Gemini Data Area Lead. Deep learning/LLM principal researcher.
Chinese name: 戴明博
1K Followers 52 FollowingThe biggest and best podcast for Duke Basketball!! Led by @blazindw and @jasondukeevans. Subscribe wherever you find your favorite podcasts!
6K Followers 121 FollowingOfficial account for SB Nation's Duke Basketball Report, covering Duke Basketball, Duke Football, and other sports in the Gothic Wonderland. #GoDuke #DukeGang
7K Followers 2K Followingtrying hard. streaming kernel dev - https://t.co/aO5X4jXGQD
blogging about life, art & systems programming. non-expert
work: @ableton, ex @trailofbits
20K Followers 325 FollowingBuilding AI that learns by interacting with the world. Associate Professor @ MIT, leading the Scene Representation Group (https://t.co/h5gvhLYZj4).