Valentine Svensson
@nxn.se
870 followers 80 following 94 posts
Molecular cell biology and stats / ML, mostly scRNA-seq. Principal Scientist at Tahoe Tx. https://www.nxn.se/archive
Posts Media Videos Starter Packs
nxn.se
.. to get connected regions*
nxn.se
In essence, it's clustering penalized by 'segmentation' (or more continuous spatial pairwise distances). You weigh how much of pairwise similarity between points has contribution from measurement (genes) or spatial location connected regions with different colors.
nxn.se
Trade relations are a lot more interesting when you're not in 5th grade :)
Reposted by Valentine Svensson
arxiv-stat-ml.bsky.social
Valentine Svensson
Negative binomial regression and inference using a pre-trained transformer
https://arxiv.org/abs/2508.04111
Reposted by Valentine Svensson
scverse.bsky.social
🎉 scverse conference 2025 Registration & Call for Abstracts NOW OPEN! 🎉
We're excited to announce that registration and the call for abstracts are officially open for the scverse Conference 2025!
Details in thread!
🧵 1/3
nxn.se
(1) Fitting hierarchical models in genetics, (2) A Stan model that runs faster with 400,000 latent parameters, (3) Super-scalable penalized maximum likelihood inference for biome problems, (4) “In the end, I basically gave up working on biology bec... - statmodeling.stat.columbia.edu/2025/07/31/b...
(1) Fitting hierarchical models in genetics, (2) A Stan model that runs faster with 400,000 latent parameters, (3) Super-scalable penalized maximum likelihood inference for biome problems, (4) “In th...
statmodeling.stat.columbia.edu