Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2510.10968 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912801276035072 |
|---|---|
| author | Zheng, Hongkai Wang, Austin Wu, Zihui Huang, Zhengyu Baptista, Ricardo Yue, Yisong |
| author_facet | Zheng, Hongkai Wang, Austin Wu, Zihui Huang, Zhengyu Baptista, Ricardo Yue, Yisong |
| contents | Derivative-free Bayesian inversion is an important task in many science and engineering applications, particularly when computing the forward model derivative is computationally and practically challenging. In this paper, we introduce Blade, which can produce accurate and well-calibrated posteriors for Bayesian inversion using an ensemble of interacting particles. Blade leverages powerful data-driven priors based on diffusion models, and can handle nonlinear forward models that permit only black-box access (i.e., derivative-free). Theoretically, we establish a non-asymptotic convergence analysis to characterize the effects of forward model and prior estimation errors. Empirically, Blade achieves superior performance compared to existing derivative-free Bayesian inversion methods on various inverse problems, including challenging highly nonlinear fluid dynamics. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_10968 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Blade: A Derivative-free Bayesian Inversion Method using Diffusion Priors Zheng, Hongkai Wang, Austin Wu, Zihui Huang, Zhengyu Baptista, Ricardo Yue, Yisong Machine Learning Derivative-free Bayesian inversion is an important task in many science and engineering applications, particularly when computing the forward model derivative is computationally and practically challenging. In this paper, we introduce Blade, which can produce accurate and well-calibrated posteriors for Bayesian inversion using an ensemble of interacting particles. Blade leverages powerful data-driven priors based on diffusion models, and can handle nonlinear forward models that permit only black-box access (i.e., derivative-free). Theoretically, we establish a non-asymptotic convergence analysis to characterize the effects of forward model and prior estimation errors. Empirically, Blade achieves superior performance compared to existing derivative-free Bayesian inversion methods on various inverse problems, including challenging highly nonlinear fluid dynamics. |
| title | Blade: A Derivative-free Bayesian Inversion Method using Diffusion Priors |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2510.10968 |