Factorizable Normalizing Flows for parameter-dependent density morphing
Davide Valsecchi 1 \orcid 0000-0001-8587-8266, Mauro Donegà 1 \orcid 0000-0001-9830-0412, Rainer Wallny 1 \orcid 0000-0001-8038-1613 · 2026
Read Factorizable Normalizing Flows for parameter-dependent density morphing by Davide Valsecchi 1 \orcid 0000-0001-8587-8266, Mauro Donegà 1 \orcid 0000-0001-9830-0412, Rainer Wallny 1 \orcid 0000-0001-8038-1613 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-30489.txt.
Description
Davide Valsecchi, Mauro Donegà, and Rainer Wallny teach a density model to morph smoothly as physical parameters change. Normalizing flows model one fixed probability density very well. Many science problems need how it changes with parameters instead. Training a separate flow per setting explodes in count. Their Factorizable Normalizing Flows sidestep that explosion.
A fixed reference flow handles the base case. A learnable change, polynomial in the parameters, deforms it. That change factorizes over parameters, which is the key move. Each knob's effect is learned from samples varying it alone. Combined responses are recovered by adding at use time. The joint space never needs sampling at all. A controlled test with two known warps validates the idea. The learned map reproduces the true warps and best likelihood. Optional interaction terms soak up leftover correlations. Cost grows in step with the number of knobs.
Catalog record
| CREATOR | Davide Valsecchi 1 \orcid 0000-0001-8587-8266, Mauro Donegà 1 \orcid 0000-0001-9830-0412, Rainer Wallny 1 \orcid 0000-0001-8038-1613 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.30489