Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Bayesian Tensor Decomposition with Diffusion Model Prior

About

Low-rank tensor decomposition (TD) is usually effective on clean, fully observed data, but it often degrades under severe missingness or noise. Low-rankness is itself a useful but limited structural prior, and additional handcrafted priors (e.g., sparsity or smoothness) still fall short of capturing the rich statistics of real-world data. To compensate for this weak inductive bias under heavy corruption, one would like to inject a learned, data-driven prior; however, the state-of-the-art diffusion models are not readily compatible with current TD and tractable posterior inference. To address these challenges, we introduce DiffBCP, a hybrid-prior Bayesian CP decomposition framework that couples a cumulative shrinkage process prior over the CP factors for automatic rank selection with an off-the-shelf pre-trained diffusion model as an implicit data prior on the reconstructed tensor. To make posterior inference tractable despite the coupling among the likelihood, low-rank constraint, and diffusion prior, we develop a split Gibbs sampler: CP factors admit conjugate updates, while the diffusion block is sampled via low-rank-guided denoising. A noise-adaptive coupling schedule further reduces sensitivity to hand-tuned annealing. Experiments on image inpainting and denoising, including high-resolution out-of-distribution images, show consistent gains over Bayesian, nonlinear, and plug-and-play TD baselines.

Zerui Tao, Qibin Zhao• 2026

Related benchmarks

TaskDatasetResultRank
Image ReconstructionImageNet
PSNR28.95
96
Image ReconstructionFFHQ
PSNR32.13
28
InpaintingFFHQ Stripe mask
PSNR27.91
5
InpaintingFFHQ Uniform 0.7 mask
PSNR32.13
5
InpaintingFFHQ Uniform 0.9 mask
PSNR28.28
5
InpaintingFFHQ Irregular mask
PSNR30.34
5
High-resolution Image RecoveryMarseille Uniform 0.9
PSNR20.15
3
High-resolution Image RecoveryMarseille Uniform 0.95
PSNR18.01
3
High-resolution Image RecoveryMarseille Irregular
PSNR20.94
3
High-resolution Image RecoveryTokyo Uniform 0.9
PSNR20.66
3
Showing 10 of 15 rows

Other info

Follow for update