Jannis Chemseddine — Generative Flows from 1D Processes and Adapting Noise to Data
2026 Invited Talk
in
Workshop: Deep Generative Model in Machine Learning: Theory, Principle and Efficacy （2nd Workshop）
in
Workshop: Deep Generative Model in Machine Learning: Theory, Principle and Efficacy （2nd Workshop）
Abstract
The aim of the talk is twofold. First, interesting noising processes besides Brownian motion are examined: the physics-inspired Kac process and the MMD gradient flow, leading to compactly supported measure curves and a better regularity of the flow matching velocity field. Second, a "quantile toolbox" for building generative models is presented: a unifying theory and a practical toolkit that turns latent noise selection into a data-driven design element.
Video
Chat is not available.
Successful Page Load