Normalize academics talking about their failures. I'll go first: We got rejected at Berghain on Saturday night.

I'm in Oslo this week for WATOC, hit me up if you wanna have a coffee ☕.

Post on previous work: and

9/ Check it yourself:. 🔗: 📄: Thanks to Khaled, Winnie, Oliver, Klaus, and Shin for the cool collaboration as well as @bifoldberlin @ml_tuberlin @TUBerlin @RIKEN_AIP_EN @RIKEN_AIP.

8/ Takeaway: Exploding TV isn’t needed. Control TV + SNR separately for faster, better sampling. Method generalizes across domains (molecules, images).

7/ Why it works? Our empirical analysis shows:. 1. Straight trajectories near data (t ≈ 0) are important (see in the inset plot).2. Broad support of pₜ(𝐱) early on → robust to errors (note how SMLD goes from small to huge range instead of staying the same)

6/ Images: Matches EDM with uniform grid. No fancy time grids like in EDM needed! VP-ISSNR on CIFAR-10/FFHQ ≈ EDM but with fewer hyperparameters!.

5/ Molecules in 8 Steps:. VP-ISSNR achieves 74% stability with 8 steps, 95% with 64 (SDE). Beats all baselines!

4/ We propose a new VP schedule 📈:. Exponential inverse sigmoid SNR (ISSNR)→ rapid decay at start/end. Generalizes Optimal Transport Flow Matching.

3/ VP variants improve existing schedules:. Take SMLD/EDM (exploding TV) → force TV=1. Result: +30% stability for molecules with 8 steps . (x-axis is NFE=number of function evals).

2/ Most schedules (like EDM by Karras or SMLD (@DrYangSong, @StefanoErmon) let TV explode (VE=variance exploding). We show constant TV (variance preserving, VP) + optimized SNR works better (ISSNR)!. (it's a wild table, sorry, but notice our VP variants I circled)

1/ Problem: Diffusion models are slow due to repeated evals but reducing steps hurts quality if the noise schedule isn’t optimal. Other schedules passively adjust variance. Can we do better?. 🔑Insight: Control Total Variance (TV) and signal-to-noise-ratio (SNR) independently!

We have a new paper on diffusion!📄. Faster diffusion models with total variance/signal-to-noise ratio disentanglement! ⚡️. Our new work shows how to generate stable molecules in sometimes as little 8 steps and match EDM’s image quality with a uniform time grid. 🧵

RT @cecamEvents: We are already on Day 3 of the workshop "Density Functional Theory and Artificial Intelligence learning from each other" i….

RT @lciernik: If two models are more similar to each other than a third on ImageNet, will this hold for medical/ satellite images? Our prep….

RT @FarnoushRJ: 🎉 Thrilled to announce that our paper MambaLRP has been accepted at #NeurIPS2024! .Explain your Mamba models in a faithful….

RT @girlbossmoder: Equivariant daughter vs augmented son.

who's going to win the election? | what's your paradigm: equivariance constraints or data augmentation?.

RT @adr_hill: Are you curious what automatic differentiation looks like in Julia? The talk @giomdal and I gave at @JuliaConOrg is finally….

I'm looking at my 2022 review of The Vegetarian written by today's Literature Nobel laureate Han Kang. I somehow twerpily disliked it. I'll re-read it but it's a different experience knowing 'who' the writer is now.

I call it now: Jurgen Schmidhuber will get the Nobel Peace prize.