Bayesian Power Steering: An Effective Approach for Domain Adaptation of Diffusion Models

Ding Huang, Ting Li, Jian Huang
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:19904-19928, 2024.

Abstract

We propose a Bayesian framework for fine-tuning large diffusion models with a novel network structure called Bayesian Power Steering (BPS). We clarify the meaning behind adaptation from a large probability space to a small probability space and explore the task of fine-tuning pre-trained models using learnable modules from a Bayesian perspective. BPS extracts task-specific knowledge from a pre-trained model’s learned prior distribution. It efficiently leverages large diffusion models, differentially intervening different hidden features with a head-heavy and foot-light configuration. Experiments highlight the superiority of BPS over contemporary methods across a range of tasks even with limited amount of data. Notably, BPS attains an FID score of 10.49 under the sketch condition on the COCO17 dataset.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-huang24l, title = {{B}ayesian Power Steering: An Effective Approach for Domain Adaptation of Diffusion Models}, author = {Huang, Ding and Li, Ting and Huang, Jian}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {19904--19928}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/huang24l/huang24l.pdf}, url = {https://proceedings.mlr.press/v235/huang24l.html}, abstract = {We propose a Bayesian framework for fine-tuning large diffusion models with a novel network structure called Bayesian Power Steering (BPS). We clarify the meaning behind adaptation from a large probability space to a small probability space and explore the task of fine-tuning pre-trained models using learnable modules from a Bayesian perspective. BPS extracts task-specific knowledge from a pre-trained model’s learned prior distribution. It efficiently leverages large diffusion models, differentially intervening different hidden features with a head-heavy and foot-light configuration. Experiments highlight the superiority of BPS over contemporary methods across a range of tasks even with limited amount of data. Notably, BPS attains an FID score of 10.49 under the sketch condition on the COCO17 dataset.} }
Endnote
%0 Conference Paper %T Bayesian Power Steering: An Effective Approach for Domain Adaptation of Diffusion Models %A Ding Huang %A Ting Li %A Jian Huang %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-huang24l %I PMLR %P 19904--19928 %U https://proceedings.mlr.press/v235/huang24l.html %V 235 %X We propose a Bayesian framework for fine-tuning large diffusion models with a novel network structure called Bayesian Power Steering (BPS). We clarify the meaning behind adaptation from a large probability space to a small probability space and explore the task of fine-tuning pre-trained models using learnable modules from a Bayesian perspective. BPS extracts task-specific knowledge from a pre-trained model’s learned prior distribution. It efficiently leverages large diffusion models, differentially intervening different hidden features with a head-heavy and foot-light configuration. Experiments highlight the superiority of BPS over contemporary methods across a range of tasks even with limited amount of data. Notably, BPS attains an FID score of 10.49 under the sketch condition on the COCO17 dataset.
APA
Huang, D., Li, T. & Huang, J.. (2024). Bayesian Power Steering: An Effective Approach for Domain Adaptation of Diffusion Models. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:19904-19928 Available from https://proceedings.mlr.press/v235/huang24l.html.

Related Material