ADAPTING NOISE TO DATA: GENERATIVE FLOWS FROM LEARNED 1D PROCESSES

Organizer

包承龙

Speaker

Gabriele Steidl

TU Berlin, Germany

Time

Tues., 16:00-17:00

Oct. 7, 2025

Venue

C548, Shuangqing Complex Building

ADAPTING NOISE TO DATA:

GENERATIVE FLOWS FROM

LEARNED 1D PROCESSES

We introduce a general framework for constructing generative models using one-dimensional noising processes. Beyond diffusion processes, we outline examples that demonstrate the flexibility of our approach. Motivated by this, we propose a novel frame-work in which the 1D processes themselves are learnable, achieved by parameterizing the noise distribution through quantile functions that adapt to the data. Our construction integrates seamlessly with standard objectives, including Flow Matching and consistency models. Learning quantile-based noise naturally captures heavy tails and compact supports when present. Numerical experiments highlight both the flexibility and the effectiveness of our method.

Joint work with J. Chemseddine, G. Kornhardt, R. Duong, P. Friz.

About the Speaker

Gabriele Steidel is a Professor at the Department of Mathematics at the Technische Universität Berlin. Her research is centered on Applied and Computational Harmonic Analysis, Convex Analysis and Optimization, Image Processing, and Machine Learning.