PulseAugur
EN
LIVE 12:05:09

New noise schedule and EM algorithm enhance diffusion model performance

By PulseAugur Editorial · [3 sources] ·

Researchers have developed a new noise schedule for diffusion models called Class-frequency Guided (CFRG) to improve generation quality, particularly for low-frequency classes in imbalanced datasets. This method addresses issues where low-density regions lead to inaccurate score estimations and high-frequency classes dominate the generation process. Experiments on image generation, classification, and text-to-image tasks using imbalanced datasets like CIFAR-100-LT and ImageNet-LT show substantial improvements over existing methods. Separately, another research paper introduces EMDiffusion, an expectation-maximization algorithm that trains diffusion models from corrupted observations, achieving state-of-the-art results in computational imaging tasks such as inpainting, denoising, and deblurring. AI

IMPACT These advancements in diffusion model training and noise scheduling could lead to higher quality image generation and improved performance in various computer vision tasks.

RANK_REASON The cluster contains two distinct research papers published on arXiv detailing novel methods for improving diffusion models.

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 3 sources. How we write summaries →

New noise schedule and EM algorithm enhance diffusion model performance

COVERAGE [3]

  1. arXiv cs.AI TIER_1 English(EN) · Jiequan Cui, Beier Zhu, Qingshan Xu, Xiaojuan Qi, Bei Yu, Hanwang Zhang ·

    Class-frequency Guided Noise Schedule for Diffusion Models

    arXiv:2606.27696v1 Announce Type: cross Abstract: In this paper, we are the first to examine the correlations between class frequency and the multi-scale noise schedule within diffusion models. For score-based generative models, low-density regions often lead to inaccurately esti…

  2. arXiv cs.LG TIER_1 English(EN) · Hanwang Zhang ·

    Class-frequency Guided Noise Schedule for Diffusion Models

    In this paper, we are the first to examine the correlations between class frequency and the multi-scale noise schedule within diffusion models. For score-based generative models, low-density regions often lead to inaccurately estimated scores, thereby compromising the generation …

  3. arXiv cs.CV TIER_1 English(EN) · Weimin Bai, Yifei Wang, Wenzheng Chen, He Sun ·

    An Expectation-Maximization Algorithm for Training Clean Diffusion Models from Corrupted Observations

    arXiv:2407.01014v2 Announce Type: replace Abstract: Diffusion models excel in solving imaging inverse problems due to their ability to model complex image priors. However, their reliance on large, clean datasets for training limits their practical use where clean data is scarce. …

RELATED ENTITIES

RELATED TOPICS