提出新方法同时解决图扩散模型的反向起始与暴露偏差，提升生成质量。
Bias mitigation in graph diffusion models
- 用新设计的Langevin采样对齐前向最大扰动分布，修正反向起始偏差。
- 引入基于分数差的新得分修正机制，缓解模型固有的暴露偏差。
- 无需修改网络结构，在多模型、多数据集上均达顶尖效果，适合图生成任务研究者。
现有图扩散模型存在显著偏差问题。我们发现，大多数模型在前向扩散过程中最大扰动分布偏离标准高斯分布，而反向采样始终从标准高斯分布开始，导致反向起始偏差。结合扩散模型固有的暴露偏差，最终降低生成质量。本文提出一套综合解决方案：为缓解反向起始偏差，采用新设计的Langevin采样算法，使其与前向最大扰动分布对齐，建立新的反向起始点；为应对暴露偏差，引入基于新定义分数差的得分修正机制。该方法无需修改网络结构，已在多个模型、数据集和任务上验证，取得当前最优结果。代码已公开于 https://github.com/kunzhan/spp。
原文摘要 · Abstract (English)
Most existing graph diffusion models have significant bias problems. We observe that the forward diffusion's maximum perturbation distribution in most models deviates from the standard Gaussian distribution, while reverse sampling consistently starts from a standard Gaussian distribution, which results in a reverse-starting bias. Together with the inherent exposure bias of diffusion models, this results in degraded generation quality. This paper proposes a comprehensive approach to mitigate both biases. To mitigate reverse-starting bias, we employ a newly designed Langevin sampling algorithm to align with the forward maximum perturbation distribution, establishing a new reverse-starting point. To address the exposure bias, we introduce a score correction mechanism based on a newly defined score difference. Our approach, which requires no network modifications, is validated across multiple models, datasets, and tasks, achieving state-of-the-art results.Code is at https://github.com/kunzhan/spp
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。