It's again time to talk Canada. 🇨🇦

CS grad school and faculty app deadlines are coming up soon. If you are applying to the US, you should also be applying to Canada. The two are far more similar than different.

Ask me anything about Canada in this🧵, and I'll answer honestly. A good place to start is my thread and AMA from last year. In short, Canada and US are incredibly similar in several dimensions, including culturally and geographically. We talk everything from admission requirements to money.

https://twitter.com/thegautamkamath/status/1463185126960115721

🧵Fields medalist June Huh shares an early math experience: a chess puzzle in the game "The 11th Hour." Story and figures from nytimes.com/live/2022/07/0….

Can you swap the positions of the black and white knights? Seems hard, right? A new perspective makes it almost trivial! 1/n We're going to define a graph over the (irregular) chess board. First of all, let's number the squares to give them names. 2/n

New workshop at @icmlconf: Updatable Machine Learning (UpML 2022)!

Training models from scratch is expensive. How can we update a model post-deployment but avoid this cost? Applications to unlearning, domain shift, & more!

Deadline May 12
upml2022.github.io #ICML2022 (1/3) Ft stellar lineup of invited speakers including: Chelsea Finn (@chelseabfinn), Shafi Goldwasser, Zico Kolter (@zicokolter), Nicolas Papernot (@NicolasPapernot), & Aaron Roth (@Aaroth)
They've studied UpML in a variety of contexts, including unlearning, robustness, fairness (2/3)

🧵One of the important principles in technical communication (i.e., writing a paper, giving a talk) of complex ideas is *organization*.
That is, making it clear what the major components are & how they fit together. If you do this well you are probably 90% of the way there. (1/n) A "top down" approach is usually the tried and true method. First communicate the high-level ideas/steps, before delving into their details. This is good because it's "truncatable": at some point "down the tree" it's ok if the audience misses a step (and they know this). (2/n)

How many of your research papers do you think will be relevant a year from now? Five years from now? 100 years from now? How do you feel about your answer? Thinking back to the time in 2017 (arxiv.org/abs/1704.03866) we used a result from a 120 year old math paper written in German (degruyter.com/document/doi/1…). Though we actually trusted an English language simplification... from 1941 (projecteuclid.org/journals/bulle…).

Paper awards for @NeurIPSConf have been announced!🎉#NeurIPS2021 blog.neurips.cc/2021/11/30/ann…

Congrats to all the winners, I'll link to the Outstanding Paper Awards 🧵

1. A Universal Law of Robustness via Isoperimetry, by @SebastienBubeck & @geoishard.

arxiv.org/abs/2105.12806 (1/n) Outstanding Paper Award 2. On the Expressivity of Markov Reward, by @dabelcs, @wwdabney, @aharutyu, @Mark_Ho_, @mlittmancs, Doina Precup, and Satinder Singh.

arxiv.org/abs/2111.00876 (2/n)

New paper on arXiv: "Efficient Mean Estimation with Pure Differential Privacy via a Sum-of-Squares Exponential Mechanism," with @Samuel_BKH and @mahbodm_.

Finally resolves an open problem on my mind since April 2019 (~2.5 years ago).

arxiv.org/abs/2111.12981
🧵Thread ⬇️ 1/n We give the 1st algorithm for mean estimation which is simultaneously:
-(ε,0)-differentially private
-O(d) sample complexity
-poly time
The fact we didn't have such an algorithm before indicates something was missing in our understanding of multivariate private estimation. 2/n

(Thread) How can I say no to this? My course will be made public! Stay tuned.
I was pulling your leg here. Here's my (hot? cold?) take: academics have a *moral obligation* to make as much of their material public as possible. Not just papers, but code, talks, and lectures. 1/n

https://twitter.com/thegautamkamath/status/1303335905072222215

95% of the work for most content is preparing it. So why not present it so the whole world can appreciate it? arXiv, Github, and video sites.
All academics should learn how to record videos in high quality, say using @OBSProject (not just Zoom). It's imperative for the future 2/n