iiinova — where your digital self keeps evolving

Reposted by Paul Hagemann

Moritz Hürtgen @moritzhuertgen.de · 12/06/2026

Jede Wahrheit braucht einen Mutigen, der sie promptet

939336

Paul Hagemann @yungbayesian.bsky.social · 07/06/2026

deep down we all knew the mcmc ppl were somehow right

010

Reposted by Paul Hagemann

Aaron Roth @aaroth.bsky.social · 12/03/2026

My favorite translation use case is from informal english to precise mathematics.

092

Reposted by Paul Hagemann

Simon Olsson @smnlssn.bsky.social · 27/05/2025

We are looking for someone to join the group as a postdoc to help us with scaling implicit transfer operators. If you are interested in this, please reach out to me through email. Include CV, with publications and brief motivational statement. RTs appreciated!

1158

Paul Hagemann @yungbayesian.bsky.social · 26/05/2025

what is so misunderstood about (3)?

200

Reposted by Paul Hagemann

Simon Olsson @smnlssn.bsky.social · 06/05/2025

2025 CHAIR Structured Learning Workshop -- Apply to attend: ui.ungpd.com/Events/60bfc...

0135

Paul Hagemann @yungbayesian.bsky.social · 27/02/2025

ist auch logisch, dass greenpeace/foodwatch bspw den grünen näherstehen, da die ja die themen bespielen. neutralität wäre da ja eher lächerlich

000

Paul Hagemann @yungbayesian.bsky.social · 18/02/2025

interesting point, but i would say (true) memorization is mathematically impossible. the underlying question is what generalization means when we are given finite training samples. it depends on the model and how long you train, see proceedings.neurips.cc/paper_files/... and arxiv.org/abs/2412.20292

proceedings.neurips.cc

Score-Based Generative Models Detect Manifolds

000

Paul Hagemann @yungbayesian.bsky.social · 07/02/2025

yes i agree, but for diffusion such a constant velocity/score field does not even exist

110

Paul Hagemann @yungbayesian.bsky.social · 06/02/2025

so in diffusion models the time schedule is so that we cannot have straight paths velocity fields (i.e., v_t(x_t) is constant in time), as opposed to flow matching/rectified flows where it is possible to obtain such paths (although it requires either OT/rectifying...)

100

Paul Hagemann @yungbayesian.bsky.social · 23/01/2025

Check out our github and give it a try yourself! Lots of potential in exploring stuff like this also to other domains (medical imaging, protein/bio stuff)! github.com/annegnx/PnP-... Also credit goes to my awesome collaborators Anne Gagneux, Sego Martin and Gabriele Steidl!

github.com

100

Paul Hagemann @yungbayesian.bsky.social · 23/01/2025

Compared to diffusion methods, we can handle arbitrary latent distributions and also get (theoretically) straighter paths! We evaluate on multiple image datasets against flow matching+diffusion+standard PnP based restoration methods!

100

Paul Hagemann @yungbayesian.bsky.social · 23/01/2025

Our algorithm proceeds as follows: we do a gradient step on the data fidelity, reproject onto the flow matching path and then denoise using our flow matching model. This is super cheap to do!

100

Paul Hagemann @yungbayesian.bsky.social · 23/01/2025

Therefore, we use the plug and play framework and rewrite our velocity field (which predicts a direction) to instead denoise the image x_t (i.e., predict the MMSE image x_1). Then we obtain a "time" conditional PnP version, where we solve do the forward backward PnP at the current time and reproject

100

Paul Hagemann @yungbayesian.bsky.social · 23/01/2025

Our paper "PnP-Flow: Plug-and-Play Image Restoration with Flow Matching" has been accepted to ICLR 2025. Here a short explainer: We want to restore images (i.e., solve inverse problems) using pretrained velocity fields from flow matching. However, using change of variables is super costly.

1155

Paul Hagemann @yungbayesian.bsky.social · 01/01/2025

very nice paper, only had a quick glimpse, but another aspect is that the optimal score estimator explodes if we approach t -> 0, which NNs ofc cannot replicate. how does this influence the results?

110

Paul Hagemann @yungbayesian.bsky.social · 28/11/2024

i guess the adam paper is a pretty good indicator how much ml papers are being published. looks like we are saturating since 2021

120

Paul Hagemann @yungbayesian.bsky.social · 24/11/2024

same experience here. i am not sure we need actual conference reviewing at all. why do we not all publish on openreview and if i use your paper/build upon/read it, i can write my opinion on it? without the accept reject stamp.

000

Paul Hagemann @yungbayesian.bsky.social · 20/11/2024

Here, one can see FID results for different beta! Indeed it seems to be fruitful to restrict mass movement in Y for class conditional cifar! We apply this also to other interesting inverse problems, the article can be found at arxiv.org/abs/2403.18705

010

Paul Hagemann @yungbayesian.bsky.social · 20/11/2024

We want to approximate this distance with standard OT solvers, and therefore introduce a twisted cost function. With this at hand, we can now do OT flow matching for inverse problems! The factor beta controls how much mass leakage we allow in Y.

120

Paul Hagemann @yungbayesian.bsky.social · 20/11/2024

This object has already been of some interest, i.e., it pops up in the theory of gradient flows. It generalizes the KL property quite nicely, and unifies some ideas present in conditional generative modelling. For instance, its dual is the loss usually used in conditional wasserstein gans.

100

Paul Hagemann @yungbayesian.bsky.social · 20/11/2024

Now does the same hold for the Wasserstein distance? Unfortunately not, since moving mass in Y-direction can be more efficient for some measures. However, we can fix that if we restrict the suitable couplings to ones, that only move mass in Y-direction.

100

Paul Hagemann @yungbayesian.bsky.social · 20/11/2024

In a somewhat recent paper we introduced conditional Wasserstein Distances. They generalize a property that basically explains why KL works well for generative modelling, the chain rule of KL! It says that if one wants to approximate the posterior, one can also minimize the KL between joints.

1150

Reposted by Paul Hagemann

Marvin Schmitt @marvin-schmitt.com · 17/11/2024

I created a starter pack for simulation-based inference (aka. likelihood-free inference). Let me know if you’d like me to add you. go.bsky.app/GVnJRoK

164218