Robust Posterior Diffusion-based Sampling via Adaptive Guidance Scale

2025-11-23

AI Summary

Diffusion models as generative priors for image inverse problems face a fundamental trade-off between prior guidance strength and data fidelity: excessive guidance introduces artifacts, while insufficient guidance leads to slow convergence and suboptimal reconstructions. To address this, we propose an adaptive likelihood guidance mechanism that eliminates manual hyperparameter tuning. It dynamically adjusts the sampling step size by evaluating the consistency between intermediate likelihood gradients and observational constraints via a dual-gradient approximation. This strategy is inherently compatible with diverse diffusion schedulers and stochastic samplers, enabling robust balance across tasks such as super-resolution and deblurring. Experiments on CelebA-HQ and ImageNet-256 demonstrate significant improvements in perceptual quality (lower FID and LPIPS), outperforming existing diffusion-based inverse solvers, while maintaining high fidelity (superior PSNR and SSIM) and strong robustness across varying noise levels.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Diffusion models have recently emerged as powerful generative priors for solving inverse problems, achieving state-of-the-art results across various imaging tasks. A central challenge in this setting lies in balancing the contribution of the prior with the data fidelity term: overly aggressive likelihood updates may introduce artifacts, while conservative updates can slow convergence or yield suboptimal reconstructions. In this work, we propose an adaptive likelihood step-size strategy to guide the diffusion process for inverse-problem formulations. Specifically, we develop an observation-dependent weighting scheme based on the agreement between two different approximations of the intractable intermediate likelihood gradients, that adapts naturally to the diffusion schedule, time re-spacing, and injected stochasticity. The resulting approach, Adaptive Posterior diffusion Sampling (AdaPS), is hyperparameter-free and improves reconstruction quality across diverse imaging tasks - including super-resolution, Gaussian deblurring, and motion deblurring - on CelebA-HQ and ImageNet-256 validation sets. AdaPS consistently surpasses existing diffusion-based baselines in perceptual quality with minimal or no loss in distortion, without any task-specific tuning. Extensive ablation studies further demonstrate its robustness to the number of diffusion steps, observation noise levels, and varying stochasticity.

Problem

• Balancing prior and data fidelity in diffusion-based inverse problems
• Adaptive likelihood step-size strategy for improved reconstruction quality
• Robust sampling method for super-resolution and deblurring tasks

Innovation

• Adaptive likelihood step-size strategy
• Observation-dependent weighting scheme
• Hyperparameter-free posterior diffusion sampling