Fad of the week? Are KANs here to stay? Have Kolmogorov-Arnold Networks always been here?

Anyway huge respect for a giant like Kolmogorov

And Terminator 2 was pretty good too!

Wait until lecun quietly updates his cognitive architecture to use the ideas behind it and claim all this is just very widely known knowledge.

1/n Math Meets AI: Kolmogorov-Arnold Networks Unleash the Power of Composition

Imagine a world where deep learning models, the enigmatic engines driving the AI revolution, are no longer shrouded in mystery. What if we could peer into their inner workings, understand their

y este teorema????? como mierda no te enseñan esto en fisica???? me lo guardo

Stolen from FB

I will be attending ICLR in Vienna next week. Who else? 🙂
2 papers at the main conf:

1/2 [Wednesday, May 8, 10:45]
The work led by wonderful Mintong 'COLEP: Certifiably Robust Learning-Reasoning Conformal Prediction via Probabilistic Circuits' cc Linyi Li and Bo Li

All you need is Kolmogorov–Arnold Network! 🔥🔥🔥 complete with GitHub repo 🚀🚀🚀🚀🚀

'KAN: Kolmogorov–Arnold Networks' from Massachusetts Institute of Technology (MIT) and Caltech

h/t Matt

Kolmogorov-Arnold network obliterates Deepmind's results with much smaller networks and much more automation.

KANs also discovered new formulas for signature and discovered new relations of knot invariants in unsupervised ways. Incredible 🔥🔥🔥🔥🔥

The bell 🔔has tolled for…

Turns out, a simple sum operator might just render #deeplearning obsolete. Solid math-based ideas are proving once again to outperform gimmicks.

#kolmogorovarnoldnetwork

#kan #kolmogorovarnoldnetwork

I LOVE CP (conformal prediction)

Valeriy M., PhD, MBA, CQF Wow, this guy even maintains a list of Conformal Prediction, it's like a treasure trove. It's great to see so many people researching things beyond LLM.

Who invented Kolmogorov-Arnold theorem?

#deeplearning

Kolmogorov-Arnold strikes again!
Little known fact: this theorem features inside one of the seminal papers on permutation invariant neural nets (Deep Sets), showing an intricate connection between such representations and the way set/GNN aggregators are built (as a special case).

Valeriy M., PhD, MBA, CQF +1

I think we’ll eventually get rid of the term deep learning.

heuristics has always been a huge part of computer science - the whole big & sophisticated packaging approach did cause that 80/20 part of the art get forgotten a little bit.

3⃣ Victor Quach, Adam Fisch, Tal Schuster, YalaTweets, Jae Ho Sohn, MD, MS, along w/ #JameelClinic PIs Tommi Jaakkola & Regina Barzilay propose an extension of conformal prediction tailored specifically to generative language models. Check it out on 5/8 in Session 3: arxiv.org/abs/2306.10193

コルモゴロフ、復活！

Valeriy M., PhD, MBA, CQF Sir David Cox once visited Kolmogorov in his home in Moskva, he told me, c. 1970. He had some sort of *muscle training stretching apparatus* on his wall, and asked Cox to try it, in which he failed miserably. Kolmogorov then showed him expertly how to operate such a thing.

im thinking about trying this next time