扩散模型已成为强大的生成工具,在样本质量上可与 GAN 媲美,并能反映自回归模型的似然分数。这些模型中的一个子类(以 DDIM 为代表)表现出一种固有的不对称性:它们在 T 个步骤上进行训练,但在生成过程中仅从 T 的一个子集中采样。这种选择性采样方法虽然针对速度进行了优化,却无意中遗漏了来自未采样步骤的关键信息,从而可能损害样本质量。为解决这一问题,我们提出了 S²-DMs,这是一种采用创新 L_skip 的全新训练方法,经过精心设计以重新整合选择性采样阶段被遗漏的信息。该方法具有多方面优势:显著提升样本质量,实现极其简单,只需极少的代码修改,并且足够灵活,可兼容多种采样算法。在 CIFAR10 数据集上,使用我们的算法训练的模型在各种采样算法(DDIM、PNDM、DEIS)和不同采样步数(10、20、……、1000)下,相比使用传统方法训练的模型提升了 3.27% 至 14.06%。在 CELEBA 数据集上,提升幅度为 8.97% 至 27.08%。代码和更多资源可在 GitHub 上获取。
英文摘要:
Diffusion models have emerged as powerful generative tools, rivaling GANs in sample quality and mirroring the likelihood scores of autoregressive models. A subset of these models, exemplified by DDIMs, exhibit an inherent asymmetry: they are trained over $T$ steps but only sample from a subset of $T$ during generation. This selective sampling approach, though optimized for speed, inadvertently misses out on vital information from the unsampled steps, leading to potential compromises in sample quality. To address this issue, we present the S$^{2}$-DMs, which is a new training method by using an innovative $L_{skip}$, meticulously designed to reintegrate the information omitted during the selective sampling phase. The benefits of this approach are manifold: it notably enhances sample quality, is exceptionally simple to implement, requires minimal code modifications, and is flexible enough to be compatible with various sampling algorithms. On the CIFAR10 dataset, models trained using our algorithm showed an improvement of 3.27% to 14.06% over models trained with traditional methods across various sampling algorithms (DDIMs, PNDMs, DEIS) and different numbers of sampling steps (10, 20, ..., 1000). On the CELEBA dataset, the improvement ranged from 8.97% to 27.08%. Access to the code and additional resources is provided in the github.
Preference Alignment on Diffusion Model: A Comprehensive Survey for Image Generation and Editing
University of Liverpool(利物浦大学);Institute of Software Chinese Academy of Sciences(中国科学院软件研究所);University of Edinburgh(爱丁堡大学);University of Bristol(布里斯托大学);University of Exeter(埃克塞特大学)
Slight Corruption in Pre-training Data Makes Better Diffusion Models
Carnegie Mellon University(卡内基梅隆大学);The University of Hong Kong(香港大学);Mila - Quebec AI Institute(米拉-魁北克人工智能研究所);RIKEN AIP(理化学研究所人工智能项目);The University of Tokyo(东京大学);William & Mary(威廉玛丽学院);MBZUAI(穆罕默德·本·扎耶德人工智能大学)
Enabling Local Editing in Diffusion Models by Joint and Individual Component Analysis
National Technical University of Athens(雅典国家技术大学);Archimedes AI, Athena RC(雅典娜研究中心阿基米德人工智能);National and Kapodistrian University of Athens(雅典国立卡波季斯特里安大学);The Cyprus Institute(塞浦路斯研究所)
University of British Columbia(不列颠哥伦比亚大学);University of Tübingen(蒂宾根大学);International Max Planck Research School for Intelligent Systems (IMPRS-IS)(国际马克斯·普朗克智能系统研究院);Technical University of Denmark(丹麦技术大学)
Unraveling the Potential of Diffusion Models in Small Molecule Generation
Department of Computer Science and Engineering, University of Connecticut(计算机科学与工程系,康涅狄格大学);Department of Biomedical Engineering, University of Connecticut(生物医学工程系,康涅狄格大学)
Trustworthy Text-to-Image Diffusion Models: A Timely and Focused Survey
WMG, The University of Warwick(沃里克大学工学院);Dept. of Computer Science, University of Liverpool(利物浦大学计算机科学系);Dept. of Computer Science & Engineering, Chalmers University of Technology(楚科奇斯技术大学计算机科学与工程系);James-Watt Engineering School, University of Glasgow(格拉斯哥大学詹姆斯-沃特工程学院)
Responsible Diffusion: A Comprehensive Survey on Safety, Ethics, and Trust in Diffusion Models
School of Cyber Science and Engineering, Southeast University(电子科技大学信息与通信工程学院);Data61, CSIRO(数据61,CSIRO);Zhejiang Laboratory(浙江省实验室);school of Computer Science and Artificial Intelligence, Wuhan University of Technology(计算机科学与人工智能学院,武汉理工大学);College of Computing and Data Science, Nanyang Technological University(计算科学与数据科学学院,南洋理工大学)
Dual-domain Multi-path Self-supervised Diffusion Model for Accelerated MRI Reconstruction
Department of Radiology, Northwestern University(放射科,西北大学);Department of Biomedical Engineering, Huazhong University of Science and Technology(生物医学工程系,华中科技大学)