![[2602.20549] Sample-efficient evidence estimation of score based priors for model selection](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20549] Sample-efficient evidence estimation of score based priors for model selection
Summary
The paper presents a novel estimator for model evidence in Bayesian inverse problems, particularly using diffusion models, enhancing sample efficiency and accuracy.
Why It Matters
This research addresses the critical challenge of selecting appropriate priors in Bayesian inverse problems, which is vital for accurate model selection. The proposed method improves the efficiency of evidence estimation, making it applicable in complex scenarios like black hole imaging, thus advancing the field of machine learning and computer vision.
Key Takeaways
- Introduces a sample-efficient estimator for model evidence using diffusion priors.
- Demonstrates the method's application in ill-posed inverse problems.
- Shows empirical validation against analytically computable model evidence.
- Facilitates model selection and diagnosis of prior misfit in complex scenarios.
- Utilizes intermediate samples from posterior sampling for improved accuracy.
Computer Science > Machine Learning arXiv:2602.20549 (cs) [Submitted on 24 Feb 2026] Title:Sample-efficient evidence estimation of score based priors for model selection Authors:Frederic Wang, Katherine L. Bouman View a PDF of the paper titled Sample-efficient evidence estimation of score based priors for model selection, by Frederic Wang and 1 other authors View PDF HTML (experimental) Abstract:The choice of prior is central to solving ill-posed imaging inverse problems, making it essential to select one consistent with the measurements $y$ to avoid severe bias. In Bayesian inverse problems, this could be achieved by evaluating the model evidence $p(y \mid M)$ under different models $M$ that specify the prior and then selecting the one with the highest value. Diffusion models are the state-of-the-art approach to solving inverse problems with a data-driven prior; however, directly computing the model evidence with respect to a diffusion prior is intractable. Furthermore, most existing model evidence estimators require either many pointwise evaluations of the unnormalized prior density or an accurate clean prior score. We propose \method, an estimator of the model evidence of a diffusion prior by integrating over the time-marginals of posterior sampling methods. Our method leverages the large amount of intermediate samples naturally obtained during the reverse diffusion sampling process to obtain an accurate estimation of the model evidence using only a handful of posterior...
