Charles Margossian

Application for a postdoc research fellowship in computational mathematics at the Flatiron Institute in New York are now open!

apply.interfolio.com/173401

📆 Deadline is December 1st.

🔭 This is an excellent place to do research at the interface of ML, stats and the natural sciences.

I also like to describe this paper as a discussion on what is the best circle to approximate an ellipse :)

🧵 4/4

This paper contributes to the foundational theory of VI, and dives deep into both conceptual and practical questions such as: How do we measure uncertainty in high-dimensions? How should we measure discrepancy between probability distributions?

🧵 3/

The two main results of the paper are:
1️⃣ An impossibility theorem that shows that any factorized (mean-field) approximation of VI can at beast learn one of three measures of uncertainty
2️⃣ An ordering of divergences used as objectives for VI based on the uncertainty in their approximation.

🧵 2/

My paper with Loucas Pillaud-Vivien and Lawrence Saul, “Variational Inference for Uncertainty Quantification: An Analysis of Trade-offs”, has been accepted for publication in the Journal of Machine Learning Research.

📃 arxiv.org/abs/2403.13748

🧵 1/

Release of CmdStan 2.37 We are very happy to announce that the 2.37.0 release of CmdStan is now available on Github! As usual, the release of CmdStan is accompanied by new releases of Stan Math, core Stan, and Stanc3. Thi…

Stan v2.37 has been released!

blog.mc-stan.org/2025/09/02/r...

Yes, in principle, I start at UBC Statistics today. But right now, I'm running around the Frankfurt airport to catch my flight to Vancouver .... 🏃‍♂️🧳✈️

www.stat.ubc.ca/news/charles...

GitHub - flatironinstitute/stan-playground: Run Stan models in the browser Run Stan models in the browser. Contribute to flatironinstitute/stan-playground development by creating an account on GitHub.

💻 This was also my first time using Stan playground (github.com/flatironinst...) to teach a class! Thank you Brian Ward for creating this tool and helping me set it up for the class!

📔 My course: "Bayesian Statistics: a practical introduction." We covered Bayesian models (priors and likelihoods), Markov chain Monte Carlo and uncertainty aware cross-validation. Most of our discussion was motivated by an example from epidemiology.

Earlier this month, I taught at the summer school on "cryptography, statistics and machine learning" (mathschool.ysu.am) hosted by Yerevan State University in Armenia 🇦🇲

🙏 Thank you to the organizers for putting together such a wonderful event! I truly enjoyed interacting with the students.

👨‍💻 Credit also to Brian Ward and Steve Bronder for their contribution to the C++ implementation and integration with the Stan ecosytem. (From what I understand, WALNUTS is not part of the next Stan release but you can use it on models written in Stan!!)

New manuscript by Nawaf Bou-Rabee, Bob Carpenter, Tore Kleppe and Sifan Liu on the WALNUTS algorithm which improves of the NUTS sampler by introducing a locally adaptive step size.

📜 Paper: arxiv.org/pdf/2506.18746
💻 Code: github.com/bob-carpente...

🇸🇬 Next stop: Singapore for BayesComp'25 (bayescomp2025.sg) The organizers put together a wonderful program!

I'll be:
🪑 chairing the session on "Parallel comp for MCMC"
🎙️ speaking at the session on "Advances in VI"

Looking forward to meeting researchers and catching up with colleagues.

Research opportunity for a graduate student in ecology 🌳 at UBC 🇨🇦 with Lizzie Wolkovich and the Temporal Ecology lab (temporalecology.org).

📝 Apply here: temporalecology.org/joining-the-... by July 1st 2025!

The abstract sounds fascinating (see attached).

CmdStan & Stan 2.37 release candidate I am happy to announce that the latest release candidates of CmdStan and Stan are now available on Github! This release cycle brings the embedded Laplace approximation, a sum-to-zero matrix type, new...

🧑‍💻 Candidate release for Stan 2.37 is out: discourse.mc-stan.org/t/cmdstan-st.... Lots of exciting features to try out, including:
- embedded/integrated Laplace approximation
- new constrained types (e.g. sum_to_zero_matrix)
- built-in constraint transformations exposed

Taking our Models Seriously (my talk at StanBio Connect, this Friday 9am)
statmodeling.stat.columbia.edu/2025/05/27/t...

🙏This award is this much more meaningful to me in that it celebrates my collaboration with the amazing Lawrence Saul (users.flatironinstitute.org/~lsaul/).

💡We provide theory on VI's ability to recover certain statistics, despite misspecification---that is in settings where we do NOT drive the KL-divergence to 0.

👉 VI is provably good at recovering the mean and correlation matrix.

✨ Thank you #AISTATS for the best paper award!!

📜 arxiv.org/abs/2410.11067

💡What does VI learn and under what conditions? The answer lies in symmetry.

🤝 Honored to share this award with my co-author Lawrence Saul from @flatironinstitute.org

🇹🇭 Just arrived in Phuket, Thailand for #AISTATS 2025.

📃 I'll presenting my recent work with Lawrence Saul on Variational Inference in Location-Sacale Families: arxiv.org/abs/2410.11067

DM if you are in town and want to connect at the conference!

For the first time, Journal-to-Conference papers have entered the schedule of #AISTATS2025 🎉

To celebrate here is a bingo card: should you be among the first to meet all of our amazing Journal-to-Conference presenters (I need proof!), I’ll buy you a drink... provided that you can find me too!

📚 Putting together a reading list for #AISTATS 2025. I already found a few very good papers, and I'm curious to hear about more accepted publications.

Elections for Stan Governing Body 2025 It’s this time of the year again! (And in fact, we’re a little bit overdue.) We’re renewing the Stan Governing Body (SGB) with all 5 seats up for grabs. Current SGB members may still run, however the...

📣 Nominations for the Stan Governing Body are open for another two weeks. Calling on all members of the Stan community to consider this role.

🔗 discourse.mc-stan.org/t/elections-...

💡 A post about my own experience: statmodeling.stat.columbia.edu/2025/03/15/e...

I'm thrilled to share that, starting this summer, I will continue my academic journey as an assistant professor of statistics at the University of British Columbia in Vancouver, Canada.

Full statement here: charlesm93.github.io/files/letter...

Petition to form a Bayesian Deep Learning section at ISBA Petition to form a Bayesian Deep Learning section at ISBA We, the undersigned members of the International Society for Bayesian Analysis (ISBA), petition the ISBA board to establish a new Bayesian D...

🚀 Exciting news! We're proposing a new *Bayesian Deep Learning* section within @isba-bayesian.bsky.social .
If you support this initiative, add your name to the petition and help make it happen!
🔗 Petition: tinyurl.com/527vaamz
Sinead Williamson, Theo Papamarkou @vincefort.bsky.social Sara Wade
1/2

A longer post on the election and my own experience on the SGB: statmodeling.stat.columbia.edu/2025/03/15/e...

... as well as some thoughts trying to both be a good developer and a good PhD student.

Elections for Stan Governing Body 2025 It’s this time of the year again! (And in fact, we’re a little bit overdue.) We’re renewing the Stan Governing Body (SGB) with all 5 seats up for grabs. Current SGB members may still run, however the...

For details: discourse.mc-stan.org/t/elections-...

Also, encourage people whom you think would be good for this role.

📣 Calling all members of the Stan community to action!

❕We're renewing the Stan Governing Body. This is a fantastic way to contribute to the project. The SGB co-organizes StanCon and related events, funds developers, and helps set the directions of the project.

🧵 Link in thread.

StanBio Connect 2025 – StanBio Connect Advancing Biomedical Research with Stan

StanConnects are a series of one-day online conferences for the application of Stan to specific areas. This year, we're kicking off with Stan for biology (broadly defined), co-organized by @ericnovik.bsky.social and @vianeylb.bsky.social, on 30 May '25!!

stanbio.org

With all that said: no, I do not use the same tools as 3blue1brown. My animations are super basic (done in keynote). But we have similar color palettes.