Sam Cohen

My latest pseudo-mathematical project....

@viditnanda I will make sure my next negative review highlights the work of V Nanda.

One of the most impressive things about anonymous peer reviews is the fact that you have absolutely no idea who your reviewers are!

@viditnanda Sounds like you’ve not been paying enough attention to opportunities for ninja in the modern academic world.

For sale: Ninja sword, never used :(

Thing that would make me happy in 2025: @RockstarGames announces a Red Dead spinoff set in Australia. With inspiration from the stories of Ned Kelly, Truganini, the man from Snowy River, On our Selection, the Eureka Stockade, … there’s plenty of source material!

@viditnanda And the only sum of cubes below some n.

Please be sure to check in on your number theorist friends today: 2025 will be the only perfect square year of their lifetimes.

@keenanisalive Of course this works until you have an asymmetry which means your chain isn’t reversible… then you have to think a bit more to describe equilibrium…

We often think of an "equilibrium" as something standing still, like a scale in perfect balance. But many equilibria are dynamic, like a flowing river which is never changing—yet never standing still. These dynamic equilibria are nicely described by so-called "detailed balance"

Hey @BBCNews I know NZ is regularly left off maps, but why do you call it part of Asia?

@miniapeur Jan Obloj and Johannes Wiesel have some projects in this direction

Who are some very good researchers that use optimal transport in machine learning? Of course, I already know the obvious answers like Peyré and Cuturi.

📝Admissions for our STATML are now open! ⏱️Deadline: 8th Jan '25 STATML is an @EPSRC funded CDT in Statistics & Machine Learning. Studentships are fully funded for a 4-year PhD/DPhil research programme. 🔗For more info, visit: statml.io

@roydanroy I’m unsurprised. My general take is that these methods are only worth it for high dimensional equations, where there are few classical competitors.

Super intrigued. Anyone who's work is being called into question: care to comment?

@keenanisalive This is important as there's no guarantee that the transition matrix is diagonalizable (a usual assumption for power iteration). That's where the slightly weird Hilbert metric allows you to say things very precisely, even without that assumption!

@keenanisalive To be careful, you need to say what metric it's a contraction in - arxiv.org/abs/2309.02413

A discrete Markov chain is basically a random walk on a graph, where each outgoing edge has a fixed probability. Cool fact: any initial distribution on a sufficiently nice* Markov chain converges to its stationary distribution after many steps—because it's a contraction mapping.

@mraginsky @SimonsInstitute Very nice. Slight typo on p20 (C_P vs C_T). If I was to argue, I'd point out that in practice we often use falsifiable models (cf Box's adage) so something beyond simple falsifiability is needed as a criterion.

@NewCollegeLib @NewCollegeOx Wearing multiple crowns, and writing your name on your leg, also seem surprising fashion choices.

‘Ugh, he’s always showing off. Why can’t he just carry his sword like a normal person?’ (@NewCollegeOx, MS 65, f. 71v) #ManuscriptMonday #MedievalTwitter #Medieval

@sp_monte_carlo Then ignore my earlier comment - I was thinking of Amari-type geodesics...

@sp_monte_carlo Feels like it would be true in exponential families - what geodesic are you thinking of though?

@ChristophHeilig @ClairedeLuned @eerdmansbooks A full Bayesian analysis would also include an assessment of the reliability of SBF as a literary critic…

@ClairedeLuned @eerdmansbooks @SamCMaths A bayesian analysis would probably (haha) show that even with low priors it takes a genius to explain the work. Hence, whatever Fried seems to assume concerning culture is a hypothesis that has a low posterior in light of that evidence.

My new book is also for readers who don't like mathematics! Get it here: amzn.to/4fYWaTc @eerdmansbooks

@ChristophHeilig @eerdmansbooks Can I infer it's principally for those of us who do like mathematics?

5-year positions for early career researchers (here = PhD viva after 1st Aug 2022) in Theoretical Physics (or Mathematics, or Archaeology, Classics, Modern Languages, Politics -- 1 for each subject). Deadline 6th September asc.ox.ac.uk/post-doctoral-…

Tomorrow, Thurs 25, 11:30 a.m. CEST — 1 p.m. CEST, presenting our work at #ICML2024, "Generalization Error of Graph Neural Networks in the Mean-field Regime" Location: Hall C 4-9, Poster 508,

@sp_monte_carlo Fair point

@sp_monte_carlo The law X=0 (identically) also works, no?