Blind denoising diffusion models and the blessings of dimensionality

2026-02-10

AI Summary

This work proposes the Blind Denoising Diffusion Model (BDDM), which eliminates the reliance on explicit noise-level information during both training and sampling—a limitation inherent in conventional diffusion models. Under the assumption of low intrinsic dimensionality, theoretical analysis demonstrates that BDDM can implicitly learn an effective noise schedule and approximate the true data distribution within a polynomial number of steps. Empirical results confirm that the model accurately estimates noise variance and, by correcting the mismatch between predefined schedules and actual residual noise, achieves superior sample quality compared to non-blind counterparts. The approach is grounded in generalization theory and complexity analysis, with comprehensive validation on both synthetic and real image datasets.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Web Mining and Content Analysis: Models for Web evolution

Abstract

We analyze, theoretically and empirically, the performance of generative diffusion models based on \emph{blind denoisers}, in which the denoiser is not given the noise amplitude in either the training or sampling processes. Assuming that the data distribution has low intrinsic dimensionality, we prove that blind denoising diffusion models (BDDMs), despite not having access to the noise amplitude, \emph{automatically} track a particular \emph{implicit} noise schedule along the reverse process. Our analysis shows that BDDMs can accurately sample from the data distribution in polynomially many steps as a function of the intrinsic dimension. Empirical results corroborate these mathematical findings on both synthetic and image data, demonstrating that the noise variance is accurately estimated from the noisy image. Remarkably, we observe that schedule-free BDDMs produce samples of higher quality compared to their non-blind counterparts. We provide evidence that this performance gain arises because BDDMs correct the mismatch between the true residual noise (of the image) and the noise assumed by the schedule used in non-blind diffusion models.

Problem

• blind denoising
• diffusion models
• noise schedule
• intrinsic dimensionality
• generative modeling

Innovation

• blind denoising
• diffusion models
• implicit noise schedule
• intrinsic dimensionality
• schedule-free sampling