Dimon Lers

Unique way of shopping! A girl operates a remote-controlled car at home to buy steamed bread in Shandong Province. Due to the outbreak of novel #coronavirus, people are advised to travel less and stay at home to avoid infection.pic.twitter.com/XlS8EQLeBD

10) THE SCREENING DILEMMA: so despite 80% PCR virus test sensitivity + 90% specificity (defined above), if virus has 2% prevalence, then PPV=13.8%, meaning only 1 in 7 positive test results will be true positives! Even if virus generously has 10% prev, PPV still only 44%!pic.twitter.com/nvZ2VnYwyU

Computer science and category theory blend in the theory of "optics": different systems for viewing and interacting with data. As part of the ACT2019 school, Emily Pillmore and Mario Román have written a great overview of these: https://golem.ph.utexas.edu/category/2020/01/profunctor_optics_the_categori.html … Lenses, prisms & more!pic.twitter.com/sLcko6gGur

An agent which learned to play Mario without rewards. Instead, it was incentivized to avoid "boredom" (that is, getting into states where it can predict what will happen next). Discovered warp levels, how to defeat bosses, etc. More details: https://blog.openai.com/reinforcement-learning-with-prediction-based-rewards/ …pic.twitter.com/6ObS35iZZS

What's the observable Universe? The further you look out into the Universe, the further you look back in time! You can't see through the hot gas from 380,000 years after the Big Bang. That "wall of fire" marks the limits of the observable Universe. (2/n)pic.twitter.com/hLMS68NZ9X

Hawking computed the temperature of a black hole, and this let him compute its entropy! The entropy is proportional to the area of the event horizon. It's 1/4 of a nat per square Planck length. But a bit is ln(2) times a nat. This entropy is huge for a big black hole.pic.twitter.com/RbGRy6J2zn

A mixture of Gaussians model can be represented as a weighted points cloud (actually a measure) over the mean/covariance domain. Mixture fitting is a non-convex optimization problem. https://en.wikipedia.org/wiki/Mixture_model …pic.twitter.com/PFAesnpfko

In practice, it's a lot easier to *measure* the entropy of water than to compute it - though still tricky. Starting from there, here's how we compute the information per molecule. Physicists use "nats" of information, but we can convert nats to bits. (5/n)pic.twitter.com/7wh2Gnf3GH

A Kerr Black Hole is an uncharged black hole, which rotates about a central axis This animation shows how spacetime is distorted around a Kerr black hole by surrounding it with 3D grid ©https://youtu.be/h75BbifqDXw pic.twitter.com/QR3HY9ugJc

If you draw lines through vertices of a pentagram, and draw lines between the new points you get that way, and so on, you can get an infinity of pentagrams - all rescaled by different powers of the golden ratio! This picture was drawn by James Dolan using Mathematica. (1/n)pic.twitter.com/GCLjMGvwIu

It's good to learn stuff like trig and integration techniques before you've been seduced by logic and set theory. He might like this nice proof of the formulas for sin(α+β) and cos(α+β). It's all in the picture. I first saw this picture last year, much too late.pic.twitter.com/9osKRclhiP

Beside being useful to prove Weierstrass polynomial approximation theorem quantitatively, Bernstein polynomial approximation enjoys many desirable properties. https://en.wikipedia.org/wiki/Bernstein_polynomial …pic.twitter.com/Os4d61Egh2

Still slowly making my way through this year's NeurIPS talks. Esp like to stumble by good talks from slightly different areas, e.g. tonight liked "ML Meets Single-Cell Biology" https://slideslive.com/38921722/machine-learning-meets-singlecell-biology-insights-and-challenges … Incredible that we're mapping out cell state markov chains for tissues.pic.twitter.com/eJcCC2AGDB