尺度空间扩散
Scale Space Diffusion
把扩散时间步与尺度空间低通滤波对应起来，用下采样作为广义线性退化，让高噪声阶段只在低分辨率上做去噪，可直接用于降低高分辨率扩散采样的计算量，并提供了一种按信息量分配网络深度的设计思路。
原文依据 highly noisy diffusion states contain no more information than small, downsampled images
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论扩散模型通过噪声使图像退化，而逆转这一过程会揭示出跨时间步的信息层级。尺度空间理论则通过低通滤波呈现出类似的层级结构。我们形式化了这一联系，并证明高噪声的扩散状态所包含的信息不超过小幅下采样图像，这引出一个问题：为何它们必须在全分辨率下处理。为此，我们通过构建一族具有广义线性退化的扩散模型及其实际实现，将尺度空间融合进扩散过程。以下采样作为退化方式便得到我们提出的 \ours。为支持 \ours，我们引入 \ourmodel，这是一种 UNet 变体，仅使用网络中必要的部分即可完成保持分辨率和提升分辨率的去噪。我们在 CelebA 和 ImageNet 上评估了该框架，并分析了其在不同分辨率和网络深度下的缩放行为。我们的网站见 prateksha.github.io/projects/scale-space-diffusion。
英文原文摘要 1021 字符
Diffusion models degrade images through noise, and reversing this process reveals an information hierarchy across timesteps. Scale-space theory exhibits a similar hierarchy via low-pass filtering. We formalize this connection and show that highly noisy diffusion states contain no more information than small, downsampled images - raising the question of why they must be processed at full resolution. To address this, we fuse scale spaces into the diffusion process by formulating a family of diffusion models with generalized linear degradations and practical implementations. Using downsampling as the degradation yields our proposed \ours. To support \ours, we introduce \ourmodel , a UNet variant that performs resolution-preserving and resolution-increasing denoising using only the necessary parts of the network. We evaluate our framework on CelebA and ImageNet and analyze its scaling behavior across resolutions and network depths. Our website is available at prateksha.github.io/projects/scale-space-diffusion.