Profile avatar
stein.ke
Computer science, math, machine learning, (differential) privacy Researcher at Google DeepMind Kiwi🇳🇿 in California🇺🇸 http://stein.ke/
216 posts 3,664 followers 629 following
Regular Contributor
Active Commenter

Expectation: en.m.wikipedia.org/wiki/Horseta...

Airbnb host seemed unusually communicative and verbose, so I was starting to suspect they were using a LLM. Then I got a message saying that they are out of office from 9pm-9am but "Our system will continue monitoring messages and responding to concerns." 🤔

Here's a mundane mystery for you: My garage door has been opening by itself randomly for the past week. We have no idea what could be triggering it. It only happens every 1-2 days, which makes it hard to debug. Twice it happened at 3am, when nothing was going on.

My wife was working the night shift this week, so I had to deal with the baby overnight.🍼 Then the baby got sick, so I had to deal with him during the day too. Then the baby started to become nocturnal like mum. The week is almost over, but we are both so tired. 😩 We are going to sleep all weekend.

The word "inequality" is now blacklisted by the NSF. Mathematicians in shambles.

It's a well-known fact™ that estimating the mean of a Bernoulli distribution from n independent samples must have mean squared error Ω(1/n). But it's surprisingly difficult to find a clean proof or citation for this fact. Here's my attempt. 😅

I was randomly reminded of this amazing propaganda poster from the Great Leap Forward. (If you don't see what's wrong with the image, ask your favourite AI what's the difference between rowing and paddling.)

It's wild that I took a whole-ass class all about differentiable functions from R^2 to itself that satisfy one specific PDE.

It would be pretty funny if the retaliation to tariffs from Canada/Mexico/China targets Tesla/SpaceX/etc. and costs Elon the title of world's richest person.

"It is amazing what you can accomplish if you do not care who gets the credit." - Harry S Truman (misattributed)

The Chebyshev-Cantelli inequality is a one-sided variant of the usual Chebyshev tail inequality. It doesn't always get as much airplay, in part because the improvement is perhaps somewhat marginal, and in part because the proof is not quite as obvious as for the conventional Chebyshev.

On the difference between precision and accuracy: The prescribed dose is 7.1mL, so the pharmacy gave us a large syringe and told us to measure 7mL with it and a small syringe to measure 0.1mL.

wife: Can you make me a coffee? me: OK me: *makes coffee* me: *drinks coffee* wife: Where's my coffee? me: 😶

Good luck, everyone!

Neat trick for bounding the maximum of independent Gaussians:

The 20th Amendment says the president's term begins/ends "at noon on the 20th day of January," but does not specify a timezone. It's accepted that this means eastern time, but I like to imagine that for a few hours the east and west of the US have different presidents.

Bernstein's inequality gives concentration bounds for sums of bounded independent random variables. It takes the variance into account, so it can be tighter than Hoeffding's & Chernoff's bounds.

Binomial anti-concentration: P[ Binomial(n, k/n) = k ] ≥ Ω( 1 / √k )

The variance of X~Binomial(n,p) is E[(X-np)^2] = np(1-p). More generally, the central moments are E[(X-np)^2k] = O(knp)^k.

Wild theory for why Trump is obsessed with annexing Greenland/Canada/Panama: He wants to make a deal - Putin gets Ukraine, Xi gets Taiwan, & he gets Greenland. What a tremendous deal. War is averted. Everyone said it couldn't be done. Everyone's happy ...right?