Optimizing Noise Schedules of Generative Models in High Dimensionss

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Aranguri, Santiago, Biroli, Giulio, Mezard, Marc, Vanden-Eijnden, Eric
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913632831406080
author Aranguri, Santiago
Biroli, Giulio
Mezard, Marc
Vanden-Eijnden, Eric
author_facet Aranguri, Santiago
Biroli, Giulio
Mezard, Marc
Vanden-Eijnden, Eric
contents Recent works have shown that diffusion models can undergo phase transitions, the resolution of which is needed for accurately generating samples. This has motivated the use of different noise schedules, the two most common choices being referred to as variance preserving (VP) and variance exploding (VE). Here we revisit these schedules within the framework of stochastic interpolants. Using the Gaussian Mixture (GM) and Curie-Weiss (CW) data distributions as test case models, we first investigate the effect of the variance of the initial noise distribution and show that VP recovers the low-level feature (the distribution of each mode) but misses the high-level feature (the asymmetry between modes), whereas VE performs oppositely. We also show that this dichotomy, which happens when denoising by a constant amount in each step, can be avoided by using noise schedules specific to VP and VE that allow for the recovery of both high- and low-level features. Finally we show that these schedules yield generative models for the GM and CW model whose probability flow ODE can be discretized using $Θ_d(1)$ steps in dimension $d$ instead of the $Θ_d(\sqrt{d})$ steps required by constant denoising.
format Preprint
id arxiv_https___arxiv_org_abs_2501_00988
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Optimizing Noise Schedules of Generative Models in High Dimensionss
Aranguri, Santiago
Biroli, Giulio
Mezard, Marc
Vanden-Eijnden, Eric
Machine Learning
Recent works have shown that diffusion models can undergo phase transitions, the resolution of which is needed for accurately generating samples. This has motivated the use of different noise schedules, the two most common choices being referred to as variance preserving (VP) and variance exploding (VE). Here we revisit these schedules within the framework of stochastic interpolants. Using the Gaussian Mixture (GM) and Curie-Weiss (CW) data distributions as test case models, we first investigate the effect of the variance of the initial noise distribution and show that VP recovers the low-level feature (the distribution of each mode) but misses the high-level feature (the asymmetry between modes), whereas VE performs oppositely. We also show that this dichotomy, which happens when denoising by a constant amount in each step, can be avoided by using noise schedules specific to VP and VE that allow for the recovery of both high- and low-level features. Finally we show that these schedules yield generative models for the GM and CW model whose probability flow ODE can be discretized using $Θ_d(1)$ steps in dimension $d$ instead of the $Θ_d(\sqrt{d})$ steps required by constant denoising.
title Optimizing Noise Schedules of Generative Models in High Dimensionss
topic Machine Learning
url https://arxiv.org/abs/2501.00988