Cornell University
Optimize Your Sampling: Tuned Diffusion Sampling with Bayesian Optimization
Abstract
dc:description.abstractThe exceptional quality of diffusion based generative models has garnered a wide audience; however, the computational burden of their iterative sampling procedure can make deployment of these models cumbersome. Speeding up diffusion sampling is an area of active research that has produced new samplers, schedules, and even new models. We show that Bayesian Optimization is an efficient approach for tuning these sampling parameters and recommend its adoption as a means to optimize diffusion sampling for a specific task or objective. We show that we can achieve comparable and oftentimes better performance using less generations than alternative approaches on various image-related tasks.
Degree
thesis:*- Name thesis:degree_name
- M.S., Computer Science
- Level thesis:degree_level
- Master of Science
- Discipline thesis:degree_discipline
- Computer Science
- Grantor
- Cornell University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Zhang, Travis
- Chair dc:contributor.chair
-
- Weinberger, Kilian
- Committee member dc:contributor.committeemember
-
- Hariharan, Bharath
Subjects
dc:subject × 3Rights
dc:rights- Statement dc:rights
-
- Attribution 4.0 International
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Bibid
- 16938452
- Dc Identifier Other
-
ProQuest Submission ID: 12437
ProQuest Publication ID: 32001110