Sign in

Seth Axen 🪓

@sethaxen.com
2.7K followers 372 following 230 posts

Empowering scientists with machine learning @mlcolab.org. Sometimes #Bayesian. Usually #FOSS. Often in #JuliaLang. Expat: 🇺🇸 ➡️ 🇩🇪 💼 On the job market (remote/southwest Germany) sethaxen.com

PostsRepliesMedia
Seth Axen 🪓 @sethaxen.com · 29/09/2026
My impression is that few things are more German than complaining about Germany. I used to find this annoying. Now I think it's a good thing. It's good to have a clear idea of how things could be better and not declare everything is fine just because it's yours.
000
Seth Axen 🪓 @sethaxen.com · 29/09/2026
My first time visiting Germany from the US, I was in awe at how punctual, quiet, and smooth the trains were (looking at you SF BART). I was surprised at how critical Germans were of DB; didn't they know how good they had it? After traveling by train in Switzerland, I get it now.
110
Seth Axen 🪓 @sethaxen.com · 27/09/2026
Congrats!! 🥳
010
Seth Axen 🪓 @sethaxen.com · 21/09/2026
While intuitive, I'm pretty sure this etymology is incorrect. Sylvester coined the term "matrix" for a much less intuitive reason: Determinants were his main object of interest, so he called the thing from which determinants originated a matrix (womb): mathshistory.st-andrews.ac.uk/Miller/mathw...
mathshistory.st-andrews.ac.uk
Earliest Known Uses of Some of the Words of Mathematics (M)
130
Reposted by Seth Axen 🪓
Aki Vehtari @avehtari.bsky.social · 09/09/2026
A FAQ is what to do in case of high Pareto-k's with PSIS-LOO or PSIS-LOGO. In case of hierarchical models we can integrate out group specific parameters before PSIS. We investigated reliability of some alternative integration methods with the eventual goal of automating this arxiv.org/abs/2609.05713
Title: Approximating Bayesian leave-one-group-out cross-validation
Authors: Anna Elisabeth Riha, Svenja Jedhoff, Paul-Christian Bürkner, Aki Vehtari
Abstract: When data are grouped, hierarchical or multilevel models are commonly used to account for group-level variation with group-specific parameters. Leave-one-group-out cross-validation (LOGO-CV) is a suitable tool for evaluating predictive performance for new groups, providing an estimator of the expected log predictive density (elpd). Brute-force LOGO-CV requires one model refit per held-out group, often using computationally expensive inference algorithms such as MCMC. This is costly, particularly for large numbers of groups or complex model structures. Commonly used importance sampling approximations, intended to reduce this cost, tend to fail because the group-specific parameters of the held-out group must be integrated out. We identify two key challenges in LOGO-CV elpd estimation: approximating the LOGO posterior and computing the grouped marginal likelihood. We compare 11 strategies, including 5 newly proposed, to address them. Among others, we combine Pareto-smoothed importance sampling or adaptive importance sampling with integration techniques such as Laplace approximation, adaptive Gauss-Hermite quadrature, and bridge sampling. We evaluate these strategies in both simulation experiments and real-world case studies, which show that marginalising over the group-specific parameters substantially improves the reliability of the importance sampling approaches.
1233
Seth Axen 🪓 @sethaxen.com · 01/08/2026
Disappointing to see the witch hunt. I guess if he could be blamed for anything, it's for not adequately preparing his community to understand that using AI as a tool at any point in the process is not inherently unethical and for apologizing too quickly.
030
Seth Axen 🪓 @sethaxen.com · 23/07/2026
I wonder how they avoid falling into the trap where the prover/verifier always says the verifier/prover is wrong to get an award
000
Seth Axen 🪓 @sethaxen.com · 22/07/2026
If possible, having it or another model design experiments to attempt to falsify can be very helpful. Also asking a different model to review parts of it can work. But all of this is just to fail early; I usually need to review it all manually in the end.
000
Seth Axen 🪓 @sethaxen.com · 21/07/2026
*forgotten, blech
010
Seth Axen 🪓 @sethaxen.com · 21/07/2026
I have no idea what this means
100
Seth Axen 🪓 @sethaxen.com · 21/07/2026
Great write-up and comparison! I'd also completely forgetting about my original idea and write-up, so thanks @spinkney.bsky.social for improving it and showing it works well in practice!
270
Seth Axen 🪓 @sethaxen.com · 21/07/2026
Interestingly, I don't think my output has substantially increased. Rather, I've found that it's much lest costly now for me to tackle more difficult engineering issues in my packages, conduct associated lit reviews, and run experiments.
0143
Seth Axen 🪓 @sethaxen.com · 21/07/2026
I don't post here very often, so I think it's worth checking in here to note that since I wrote this, I've transitioned to a heavy agent-based dev workflow. Partially that's due to a huge improvement in the models, partially due to some useful skills/harnesses.
1341
Seth Axen 🪓 @sethaxen.com · 19/07/2026
For me it's usually the 50s
010
Seth Axen 🪓 @sethaxen.com · 17/07/2026
These PhD projects look like a ton of fun!
000
Seth Axen 🪓 @sethaxen.com · 17/07/2026
I still use "delve". If it was good enough for Tolkien, it's good enough for me.
030
Seth Axen 🪓 @sethaxen.com · 09/07/2026
can't. tell. if. joking. 🧐
010
Seth Axen 🪓 @sethaxen.com · 08/07/2026
Bring 👏back 🦶blue books! 📘
010
Seth Axen 🪓 @sethaxen.com · 30/06/2026
"almost" 🧐
030
Seth Axen 🪓 @sethaxen.com · 30/06/2026
To be fair, I'd never heard of "cooling centers" when I lived in CA, but almost any home or business had A/C. But in my German town, their advice for the heatwave was basically: don't bake, exercise, or eat hot food; close your shutters; take a cold shower. We don't have a mall, sooo...
110
Seth Axen 🪓 @sethaxen.com · 30/06/2026
How many times per day might an animal or plant try to kill me? Asking for a friend.
230
Seth Axen 🪓 @sethaxen.com · 26/06/2026
And yes, they were talking about Bayesian models, not Bayesian curls.
020
Seth Axen 🪓 @sethaxen.com · 26/06/2026
That feeling when you join a weekly video group check-in with your strength training coach, and the very first word you hear is "Bayesian"... 🤗
120
Seth Axen 🪓 @sethaxen.com · 21/06/2026
Not bad
000
Seth Axen 🪓 @sethaxen.com · 20/06/2026
Daredevil: Born Again has some dumb parts but is overall quite good, very much in the vein of the original Netflix show. But IMO none of Marvel's shows can compare to WandaVision. Hopefully VisionQuest doesn't suck.
010
Seth Axen 🪓 @sethaxen.com · 18/06/2026
I think this is the first one I've actually managed to solve. Got lucky on the purple: 4 × 3 100 points • 0 mistakes 🟩🌟🟩 🟨🌟🟨 🌟🟪🟪 🟦🌟🟦
000
Seth Axen 🪓 @sethaxen.com · 18/06/2026
Same. Man, I am missing AC in Germany right now.
010
Seth Axen 🪓 @sethaxen.com · 11/05/2026
who chooses to name their company "Bland"?
020
Seth Axen 🪓 @sethaxen.com · 27/04/2026
I find the smoothness of the marmoset and weasel to be unsettling.
010
Seth Axen 🪓 @sethaxen.com · 15/04/2026
Oh good, nice and easy
020
Seth Axen 🪓 @sethaxen.com · 15/04/2026
Well that's not what I needed to be thinking about today
110
Reposted by Seth Axen 🪓
TMLR Published Papers @tmlr-pub.bsky.social · 09/04/2026
Amortized Bayesian Workflow Chengkun LI, Aki Vehtari, Paul-Christian Bürkner et al. Action editor: Tom Rainforth openreview.net/forum?id=osV7adJlKD #mcmc #generative #amortized
033
Reposted by Seth Axen 🪓
ArviZ @arviz.bsky.social · 19/03/2026
ArviZ 1.0 is out! We have refactored it to be more modular, flexible & lightweight. For an overview of the changes, check the migration guide. python.arviz.org/en/stable/us...
python.arviz.org
Getting started — ArviZ 1.0.0 documentation
21510
Seth Axen 🪓 @sethaxen.com · 17/03/2026
media.tenor.com
a man in a green sweater is standing in front of a pioneer video player
ALT: a man in a green sweater is standing in front of a pioneer video player
020
Seth Axen 🪓 @sethaxen.com · 12/03/2026
Also I start to get the chills when I know I'm dealing with a special case of a hypergeometric function, because I just *know* as soon as I try differentiating or integrating something I'll be stuck with just a hypergeometric.
050
Seth Axen 🪓 @sethaxen.com · 12/03/2026
Literally every function that is non-integrable. They have evil intentions, and I don't trust them.
1100
Reposted by Seth Axen 🪓
Rafe Meager (they/them) @economeager.bsky.social · 12/03/2026
"Other extensions of the factorial function do exist, but the gamma function is the most popular and useful. " did the gamma function herself/itself write this
2638
Seth Axen 🪓 @sethaxen.com · 12/03/2026
Anyone aware of a collection of agent skills tailored towards FOSS workflows for package development and maintenance? Strongly considering writing my own.
111
Reposted by Seth Axen 🪓
Aki Vehtari @avehtari.bsky.social · 04/03/2026
If you have been using LOO-PIT, this is a must read for you! @herman-tesso.bsky.social has done excellent work with this paper! Thanks for @florencebockting.bsky.social and @aloctavodia.bsky.social for getting this to bayesplot and ArviZ. I'll notify when I have my casestudies updated with this
0162
Seth Axen 🪓 @sethaxen.com · 02/03/2026
Congrats, and this is amazing!! 😂
010
Seth Axen 🪓 @sethaxen.com · 05/02/2026
russian blue? worth. it.
010
Seth Axen 🪓 @sethaxen.com · 28/01/2026
Yes there are dozens of tasks throughout the day that I can accomplish much faster with AI, but these are usually sporadic and non-routine. Building a whole set of agents to manage these would then become its own large task with little concrete benefit that I can see.
030
Seth Axen 🪓 @sethaxen.com · 28/01/2026
Like my #1 productivity blockers are my own mental clarity/energy (which are helped by exercising, resting, and preserving a routine and focused time for thinking, writing, and hacking on stuff) and family/home duties (which AI is mostly useless for).
140
Seth Axen 🪓 @sethaxen.com · 28/01/2026
I've been seeing some folks recently explain how they're using agent-based workflows to help manage their life/work. I think it's great that people are sharing this and probably super useful for some. But I still haven't seen a use case that would be helpful for me.
141
Seth Axen 🪓 @sethaxen.com · 20/01/2026
I think @vgel.me is thinking of the subset of Pauline letters in the canon for which the scholarly consensus is that they're pseudepigraphic. Would probably include 1&2 Timothy, Titus, Ephesians, and a few others. You can find them online in any digital Bible.
150
Seth Axen 🪓 @sethaxen.com · 07/01/2026
Over the break I was picking away at a personal project where brms is clearly the best tool to use, and hot damn, is brms nice!
0101
Seth Axen 🪓 @sethaxen.com · 03/01/2026
Introduce yourself as what almost killed you: Hi there, I'm a hernia
010
Seth Axen 🪓 @sethaxen.com · 18/12/2025
taking personal branding to a whole new level
110
Reposted by Seth Axen 🪓
Demetri @phdemetri.bsky.social · 26/11/2025
I believe eigen fly. I believe eigen touch the sky.
324042
Seth Axen 🪓 @sethaxen.com · 24/11/2025
I guess there's Rogue Scholar for archiving blog posts and generating a DOI (rogue-scholar.org). The main problem is one of venue. For arXiv I'd want to rework/polish the content to increase precision, and IME that can take at least as long as writing the original blog post.
010