New paper by David Huk and Theo Damoulas! They model copulas based on the principles of diffusions and flows through processes that preserve marginal distributions but forget the dependence. Link to the paper 👉 arxiv.org/abs/2509.1....
18. arxiv.org/abs/2509.19707
'Diffusion and Flow-based Copulas'
-David Huk, Theodoros Damoulas
If a distribution has standard Gaussian marginals, then one can smoothly interpolate it towards an exact Gaussian along a path which never changes those marginals, only altering the copula.
My Pint of Science talk on bayesian statistics and its role in early phase trial design is now available on YouTube!
It was an honour to represent the Alan Turing Institute and @icr.ac.uk at #pint24 last year. I hope it’s a useful resource for those interested.
youtu.be/-471AqKp9bw?...
New paper accepted in AISTATS! David Huk (@david-huk.bsky.social), Mark Steel and Ritabrata Dutta show that any copula can be represented as a classification problem, giving a new approach to model copula densities. Link to preprint:
tinyurl.com/RatioCopula
Shreya Sinha Roy, Richard G. Everitt, Christian P. Robert, Ritabrata Dutta
Generalized Bayesian deep reinforcement learning
https://arxiv.org/abs/2412.11743
Want ⚡-fast and efficient density estimation with Bayesian properties? Quasi-Bayesian Vines are the tool you need!
We get predictive densities through recursion without any MCMC. Check out the #NeurIPS2024 poster on Friday from 4:30 to 7:30 PM in the East Exhibit Hall #3806.
tinyurl.com/53vmafc8
How do you go about taking a decision in these cases as an AC? Or if the reviewers just never reply?
(2/x)
When someone wants to become a scientist, this comes with responsibilities. They want their paper to be properly reviewed, read, and even cited. They must also contribute to keeping science as rigorous possible by writing quality reviews.
How do you conclude that a paper costs 100K EUR to the taxpayer?
It's a good way to look at things since that's mostly where academic funds come from, but it's still a very large number.
🤔Question: How do you scale recursive prediction to high dimensions?
💡Answer: with copulas!
Check out our "Quasi-Bayes meets Vines" paper, accepted at #neurips2024, and come say hi at the poster session in Vancouver!
🎥(5 min video) tinyurl.com/ycpu23cw
📄(Paper) tinyurl.com/2s4f96tm