Alex Cole

physics, computation, @AmsterdamTheory and @GRAPPAinstitute

[Alex Cole @a_e_cole]

With @bkmi13 @C_Weniger and more, we put out https://t.co/sVFTHlGZhB, our demonstration of simulation-based inference with Truncated Marginal Neural Ratio Estimation (TMNRE) for cosmology! Here's a lil' thread 🧵

[GRAPPA Centre @GRAPPAinstitute]

New GRAPPA paper: https://t.co/B5PV1asJG9 !
The authors, including @C_Weniger and former GRAPPAs @a_e_cole and Sam Witte, apply Marginal Neural Ratio Estimation to constrain astrophysical parameters describing Cosmic Dawn and Epoch of Reionization (1/2)

[datascience@uw @datascience_uw]

String theory has more solutions than the number of atoms in the observable universe. Machine learning helps Gary Shiu, @UWMadPhysics, discover patterns in incredibly vast and complex data. Read his story in Faces of Data Science: https://t.co/B6shjQXz7v @gary_shiu

[Alex Cole @a_e_cole]

@jhhalverson @KrippendorfSven Some slides (including mine) are now posted at https://t.co/meqRRyPuEV

[Sven Krippendorf @KrippendorfSven]

The ASC school on Physics Meets AI is starting with @DaniloJRezende. Great to see you all in person! #ASCschool

[Sebastian Schmon @SeBayesian]

To make simulation-based inference methods viable for imperfect simulators (simulators that only partially explain real outcomes) we need new and robust methods. We experimented with popular SBI approaches using coverage as a metric and ..oh boy 🧵

[Arnaud Delaunoy @ArnaudDelaunoy]

After two years of work, Balanced Neural Ratio Estimation (BNRE) is now released! Very happy to be part of this amazing journey! If you are willing to use BNRE for your applications, nothing more simple. Just add one line of code to your running NRE implementation.

[Gilles Louppe @glouppe]

On behalf of the organizers, I am happy to announce that the Call for Papers is now out! https://t.co/eSPKrBK7M3 Deadline: September 22. ✍️#ML4PS2022

[Gary Shiu @gary_shiu]

Such a pleasure speaking on “ML for Theoretical Physics” with @KrippendorfSven at the intellectually stimulating #HammersAndNails organized by @eilamgross . Honored to be on the scientific organizing committee alongside, among others, KyleCranmer @datascience_uw @UWMadPhysics

[Alex Cole @a_e_cole]

swyft is now published in JOSS!

[Benjamin Kurt Miller @bkmi13]

Today, we published the paper for the code of swyft, a library which implements a simulation-based inference method called Truncated Marginal Neural Ratio Estimation! https://t.co/XZXJdDKswB
@a_e_cole @C_Weniger Francesco Nattino Ou Ku and Meiert W. Grootes. @eScienceCenter

[Alex Cole @a_e_cole]

@DD_Baumann @Matteo1Biagetti @gary_shiu Indeed! On our own modest scale. Including fun stuff like RSD, HOD… here is the page with more details https://t.co/CYK0LquVSH — maybe we should call it the PNG Olympics?

[Alex Cole @a_e_cole]

Looking forward to next week in Trieste at the Focus Week "Interpretable and higher-order statistics for late-time cosmology" https://t.co/HhGkLfkIbC -- organized with @Matteo1Biagetti, @gary_shiu, and Jorge Noreña. Talks will also be broadcast via Zoom 📢🍩🧠

[Gary Shiu @gary_shiu]

Lots have been said about how large the string landscape can be. Time for an exact count. 215 billion distinct D-brane models in the work with Gregory Loges @UWMadPhysics https://t.co/I0JGpsAbng

[Sven Krippendorf @KrippendorfSven]

Don't miss the application deadline on May 31st!

https://t.co/YwOn02EG1I

[Benjamin Kurt Miller @bkmi13]

I'll present at CERN in about an hour and ten minutes. https://t.co/GAaBNuOgT6

@a_e_cole is next up in about two hours and ten minutes.

[Alex Cole @a_e_cole]

@CburgesCliff my pleasure, was great to have you visit! 🍻

[Benjamin Kurt Miller @bkmi13]

Enjoying @glouppe 's talk at Likelihood-free in Paris!

@a_e_cole, @AnauNoemi, @adamcoogan1, @C_Weniger, Kosio Karchev, Elias Dubbelman, and I present later about swyft, Truncated Marginal Neural Ratio Estimation, and applications.

[Alex Cole @a_e_cole]

@ThomasVanRiet2 from this picture i don’t think they include loop corrections

[Sven Krippendorf @KrippendorfSven]

Want to learn more about how AI and Physics connect? Come to our school in Munich the week before Oktoberfest. More details here: https://t.co/CTWMlCAC6G

@aBohrdt, @gppcarleo, @a_e_cole, @marylougab, @GittaKutyniok, @rgmelko, @DaniloJRezende

@lodepollet, @krippendorfsven

[Alex Cole @a_e_cole]

@samayanissanke @GRAPPAinstitute

[Alex Cole @a_e_cole]

@glouppe @LucaAmb @DaniloJRezende @C_Weniger It looks like in this case (hard to find actual priors in the paper, but comparing log M_* in figures 1 and 2) the likelihood is not very constraining. Posteriors fill up a good fraction of prior