Alexander Johansen
banner
alexrosejo.bsky.social
Alexander Johansen
@alexrosejo.bsky.social
Stanford CS PhD student | Recovering deep learning practitioner, now doing proofs instead of parameter sweeps. Also I like birds
Reposted by Alexander Johansen
Anupama Sridhar, Alexander Johansen
Convergence of Adam in Deep ReLU Networks via Directional Complexity and Kakeya Bounds
https://arxiv.org/abs/2505.15013
May 22, 2025 at 5:41 AM
Reposted by Alexander Johansen
A Kakeya set is the smallest space a cat can spin in every direction.

That’s your ReLU network.
Track those spins, and you get tighter control than PAC-Bayes.

Cats don’t take random walks. Neither should your optimizer.

#CatsOfML #Kakeya #CSTheory
May 20, 2025 at 7:24 PM
ChatGPT won't ever run anesthesia during surgery. Eat your spinach and do your bounds. Also, TD(0) converges for non linear functions arxiv.org/pdf/2502.05706
arxiv.org
May 20, 2025 at 7:48 PM