Vinayak Pathak

Simplifying Adversarially Robust PAC Learning with Tolerance Adversarially robust PAC learning has proved to be challenging, with the currently best known learners [Montasser et al., 2021a] relying on improper methods based on intricate compression schemes, res...

In the off chance that you didn't notice my new paper with Hassan Ashtiani and Ruth Urner among the 265 papers that were announced on cs.LG today, here it is: arxiv.org/abs/2502.07232

For 1d mean estimation when all we know is that variance is finite, does the median of means algorithm need to know an upper bound on the variance to get its guarantees? If yes, is there an algorithm that doesn't need to know this?

Interesting. I thought people had completely moved to ChatGPT/claude for this sort of stuff

I was surprised to see that 4th floor math/physics section has some very good textbooks!

What an honour. Congrats

Theoretical CS has it seems

Let’s tag @cosenal.bsky.social so he doesn’t miss out