Abstract:In this paper, we point out that suboptimal noise-data mapping leads to slow training of diffusion models. During diffusion training, current methods diffuse each image across the entire noise space, resulting in a mixture of all images at every point in the noise layer. We emphasize that this random mixture of noise-data mapping complicates the optimization of the denoising function in diffusion models. Drawing inspiration from the immiscibility phenomenon in physics, we propose Immiscible Diffusion, a simple and effective method to improve the random mixture of noise-data mapping. In physics, miscibility can vary according to various intermolecular forces. Thus, immiscibility means that the mixing of molecular sources is distinguishable. Inspired by this concept, we propose an assignment-then-diffusion training strategy to achieve Immiscible Diffusion. As one example, prior to diffusing the image data into noise, we assign diffusion target noise for the image data by minimizing the total image-noise pair distance in a mini-batch. The assignment functions analogously to external forces to expel the diffuse-able areas of images, thus mitigating the inherent difficulties in diffusion training. Our approach is remarkably simple, requiring only one line of code to restrict the diffuse-able area for each image while preserving the Gaussian distribution of noise. In this way, each image is preferably projected to nearby noise. Experiments demonstrate that our method can achieve up to 3x faster training for unconditional Consistency Models on the CIFAR dataset, as well as for DDIM and Stable Diffusion on CelebA and ImageNet dataset, and in class-conditional training and fine-tuning. In addition, we conducted a thorough analysis that sheds light on how it improves diffusion training speed while improving fidelity. The code is available at <a class="link-external link-https" href="https://yhli123.github.io/immiscible-diffusion" rel="external noopener nofollow">this https URL</a>

Multi-Class Error-Diffusion with Blue-Noise Property and Its Application

Multi-Step Denoising Scheduled Sampling: Towards Alleviating Exposure Bias for Diffusion Models

A Three-Dimensional Error-Diffusion Algorithm For Importance Sampling With Blue-Noise Property

Structure-Aware Error-Diffusion Approach Using Entropy-Constrained Threshold Modulation

Improved method of image color transfer.

Color error diffusion based on multilevel halftone

Wasserstein Blue Noise Sampling

Bilateral Blue Noise Sampling

A Noise-Model-Free Hyperspectral Image Denoising Method Based on Diffusion Model.

Channel Selection for Multispectral Color Imaging Using Binary Differential Evolution.

A Diffusion-Based Framework for Multi-Class Anomaly Detection

Blue noise for diffusion models

Your Diffusion Model is Secretly a Noise Classifier and Benefits from Contrastive Training

Enhancing Sample Generation of Diffusion Models using Noise Level Correction

Image Vectorization Using Blue-Noise Sampling

Diffusion Models With Learned Adaptive Noise

QNCD: Quantization Noise Correction for Diffusion Models

Immiscible Diffusion: Accelerating Diffusion Training with Noise Assignment

Exploring the Design Space of Diffusion Bridge Models via Stochasticity Control

NoiseDiffusion: Correcting Noise for Image Interpolation with Diffusion Models beyond Spherical Linear Interpolation

CDDM: Channel Denoising Diffusion Models for Wireless Semantic Communications