Radu Grigore

cs & math

[Radu Grigore @rgrig]

I once took a personality test. It said I'm in top 5% of disagreeable people. That means that if I agree with something I remain quiet, and if I disagree I speak. So, if some time I annoy you by disagreeing with you, think of all those times I said nothing!

[Radu Grigore @rgrig]

@HerrDreyer @BenardDreyer Condolences.

[Radu Grigore @rgrig]

nice proof

[Peyman Milanfar @docmilanfar]

A surprising & little-known results in classical statistics:

mean (μ) and median (m) are within one std deviation:

|μ−m| ≤ σ

For unimodal densities, bound is even tighter

|μ−m| ≤ 0.7746 σ

This beautiful results first appeared in a 1932 paper by Hotelling & Solomons

1/3

[Google DeepMind @GoogleDeepMind]

We’re presenting the first AI to solve International Mathematical Olympiad problems at a silver medalist level.🥈

It combines AlphaProof, a new breakthrough model for formal reasoning, and AlphaGeometry 2, an improved version of our previous system. 🧵 https://t.co/SYaLPSbIyj

[Radu Grigore @rgrig]

@ccanonne_ @SimonsInstitute the other day i randomly decide to mention that i'm fighting with re-proving a 50y old result
and the next day you randomly decide to post a picture with the guy who did the original proof 50y ago ...

[Radu Grigore @rgrig]

@ccanonne_ Maybe I should just go read the proof. But I refuse to be defeated but this damn old problem. :P

[Radu Grigore @rgrig]

@ccanonne_ With a programming job I get to do maths rarely. For the past 2 months I spent about 15min every 3 days to try to prove a theorem, and failed. (That's about 5h in total.)

This is a result from the 70s about caches. So, yeah, quite humbled.

[Radu Grigore @rgrig]

@wicko3 @PerhaloP @rlepigre or in python

[Radu Grigore @rgrig]

not spoiler: the answer is no
:)

[Radu Grigore @rgrig]

what i did today in the shower

[Joel David Hamkins @JDHamkins]

Another nice problem from my daughter's HS math club. We have finitely many switches in a row, each in position up, down, left, or right. Whenever three adjacent switches have all different positions, we can change all three to the fourth position. Can it go forever?

[Radu Grigore @rgrig]

@crankycatcoder @krismicinski K&R is a really good book

[Radu Grigore @rgrig]

@JDHamkins Oh: and, to be clear, in the programming problem i think you were told that it *is* possible -- you just had to find one way to do it.

[Radu Grigore @rgrig]

@JDHamkins I first saw this given as a programming problem in an ICPC contest. From replies, I see it's even older.

Anyway, the solution I know is the one where you look at the sum of the lengths of all segments.

[Joel David Hamkins @JDHamkins]

My daughter had a nice problem in her high-school math club. Suppose you have 1000 white points and 1000 black points in the plane, no three collinear. Can you draw segments connecting them in pairs, from white to black, using each point just once, without any edges crossing?

[Stefan Marr @smarr]

Heard we (@UniKentComp)'ve 4 PhD scholarships. No details yet, but it's likely "short notice" for applications.

If you know someone who's into Programming Languages and is interested in going deep, check our list of suggested topics: https://t.co/GypQ7lS60p

#phd #funding

[Radu Grigore @rgrig]

@smarr fairly sure that if you spend 2 more days you can come up with a query that takes only 1h

[Radu Grigore @rgrig]

@krismicinski the writing is on the wall!

[Radu Grigore @rgrig]

@ccanonne_ @thomasahle > "well, if that's the easy part ..."

what if it isn't?
they still solved an incredibly hard problem

[Radu Grigore @rgrig]

@Barteks2x @wicko3 @HerrDreyer @h_poncedeleon @poplconf One possibility: 'or 1' loads and writes one byte only.

or 1: reads 0 byte
dec 1: reads 0 int
dec 1: writes -1 int
or 1: writes 1 byte

result: -255 int

[Radu Grigore @rgrig]

@karger Thanks for the answer.

It is indeed not uniform: I tried a square with a diagonal. I lost the paper now, but I recall it was *close* to uniform.

I can do more experiments. But making "close" more precise seems like a job for theory.

[Radu Grigore @rgrig]

Since I used @karger's mincut algo in the last day of #AdventOfCode, I wondered what happens if you use that algo to sample spanning trees, by doing one more step until all selected edges make a tree.

I know the sampling is not uniform, but I guess it should be close to uniform.

[Radu Grigore @rgrig]

@krismicinski seems to be an often-rediscovered law:
https://t.co/XzeSoYriKx