Search papers, labs, and topics across Lattice.
This paper introduces a novel estimator for the model evidence of diffusion priors, addressing the challenge of prior selection in Bayesian inverse problems. The method integrates over the time-marginals of posterior sampling methods, leveraging intermediate samples from reverse diffusion to estimate model evidence. The proposed estimator, \method, achieves accurate model evidence estimation with only a small number of posterior samples and demonstrates its effectiveness in selecting appropriate diffusion model priors and diagnosing prior misfit in ill-conditioned inverse problems, including black hole imaging.
Accurately select diffusion model priors with just 20 samples, even for real-world black hole imaging, using a novel estimator that integrates over posterior sampling time-marginals.
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 samples (e.g., 20). We also demonstrate how to implement our estimator in tandem with recent diffusion posterior sampling methods. Empirically, our estimator matches the model evidence when it can be computed analytically, and it is able to both select the correct diffusion model prior and diagnose prior misfit under different highly ill-conditioned, non-linear inverse problems, including a real-world black hole imaging problem.
