Sabitlenmiş Tweet
Just released imcmc so you can make your images into probability distributions and then into gifs: github.com/ColCarroll/imc…
GIF
English
Colin Carroll
2.3K posts
Colin Carroll
@colindcarroll
Software engineer @Google. Interested in Python, math, and machine learning. Not opposed to hiking, running, and biking. PyMC3 + ArviZ developer. he/his.
Cambridge, MA Katılım Aralık 2013
859 Takip Edilen2.8K Takipçiler
Colin Carroll retweetledi
🎉 Excited to share that my book is part of @PacktPublishing’s October Sale, with up to 25% off on Amazon! 📚 Don’t miss the chance to grab a copy—sale ends Nov 1st! 👉 [packt.link/eoiCp]
English
@Ferrum_of_omega @AnthropicAI Congrats, Sam! Can't wait to see what you do!
English
Fun moment to announce I've joined @AnthropicAI on the MechInterp Team! 🚀
Anthropic@AnthropicAI
Introducing an upgraded Claude 3.5 Sonnet, and a new model, Claude 3.5 Haiku. We’re also introducing a new capability in beta: computer use. Developers can now direct Claude to use computers the way people do—by looking at a screen, moving a cursor, clicking, and typing text.
English
@tophtucker I also like thinking about how functions don't (usually) commute with integration/expectation: the most likely draw from a 10,000-d normal is all zeros, but the expected norm of a draw is 100 (and the variance is 0.5, so it is vanishingly rare to find a draw less than, say, 95).
English
@tophtucker Optimization is also different from integration (say, over possible outcomes). The "most likely" prob. for an unfair coin is 0.5, both before seeing evidence, and after seeing 1M heads and 1M tails.
English
asked @rsnous about "decisions made by nobody", the feeling that locating a decision in an organization is like locating a soul in a body, and he suggested "War and Peace". wow! now i gotta read that!
San Francisco, CA 🇺🇸 English
my own ridge_map, and @arviz_devs
Pawel Jastrzebski 🇵🇱 🇬🇧@pawjast
If you ever wondered if learning @matplotlib is worth your while - check out this list of the tools that build on top of Matplotlib or extend its capability. Translation: Learning it will most certainly pay big dividends as you dive into the wider Python ecosystem. Guaranteed.
English
@RingoHosp Some legends up there. I got to take a Galois Theory elective with him. He had an incredible (obvious?) way of getting questions: making eye contact with one person at a time and staying quiet.
English
@colindcarroll my neighbor! go purple cows. i always think i should audit some math class up there…
English
The incredible Thomas Garrity, whose book "all the math you missed" remains a favorite.
Looks like this video is on YouTube, titled "on mathematical maturity".
Knowledge Bank@xKnowledgeBANK
Very strong opening to a math video…
English
Colin Carroll retweetledi
Dear colleagues, I'm in the last year of my postdoc at @FlatironInst and I'm looking for research and faculty positions in statistics, data science, and machine learning (for the summer/fall of '25).
DM me if you know of places which would be a good fit!
English
Colin Carroll retweetledi
Imagine being able to accurately forecast air quality & climate changes. Our Bayesian Neural Fields (BayesNF) AI model melds pattern-recognition from deep learning w/ statistical rigor from Bayesian methods — a crucial step in making this a reality. More: nature.com/articles/s4146…
English
Excited that our paper "Scalable spatiotemporal prediction with Bayesian neural fields" (led by Feras Saad) was just published: nature.com/articles/s4146…
There is also a python library at github.com/google/bayesnf - give it a try if you've got some spatiotemporal problems!
English
@sp_monte_carlo Totally! A tight integration would be neat, but I'm not sure these populations talk that often (meaning: I've never talked to anyone implementing quadrature packages). I *have* struggled to reparameterize a model, realized it was only 2 dimensions, and just... made a grid.
English
I don't know any PPLs that provide easy support for quadrature. Seems useful to have an option to switch from MCMC in, say, 5 or fewer dimensions.
English
@sethaxen True, but we make do for the sake of variational inference!
English
@colindcarroll Yeah just seems like it would be tricky to adapt all the downstream analyses to work with quadrature. MC draws are a really useful intermediate representation.
English
I guess github.com/jax-ml/bayeux is getting a neighbor!
Dan F-M@exoplaneteer
🔖 Update your bookmarks! Soon github.com/google/jax will move to github.com/jax-ml/jax See the announcement here: github.com/google/jax/dis… One thing I wanted to emphasize: despite the name, this doesn't change the fact that JAX is still useful for a LOT more than just ML!!
English
Colin Carroll retweetledi
Some personal news: I'm looking for a job!
tomaugspurger.net/posts/leaving-…
Some of my (very talented) former teammates are also looking for a job, so reach out if you're hiring!
English
@Bojan93112526 @PredragGruevski Is this... few shot learning?
English
@PredragGruevski Sorry to move the goal posts, but can it get wasted the day after the competition?
English
As a former IMO competitor, I will vouch that "AI solves 4/6 problems at IMO" is a mind-boggling statement 🤯
I look forward to reading the paper with all the details. Assuming it all checks out, my hat is off to you! 🎩🙇♂️
deepmind.google/discover/blog/…
English
English
@AdrienCorenflos @sp_monte_carlo I ran into some mushroom foragers in New Hampshire last weekend. They seemed happy...
English