让扩散模型跳过零值，高效生成稀疏数据
Skipping the Zeros in Diffusion Models for Sparse Data Generation
- 只建模非零值，跳过零点节省计算
- 在物理与生物数据上质量不降反升
- 适合处理含大量零值的科学数据
扩散模型（DMs）在密集连续数据上表现优异，但未针对稀疏连续数据设计，无法准确建模代表信号缺失的精确零值。这导致其会破坏稀疏模式，并在多数为零的区域进行无意义计算。本文提出稀疏性利用扩散模型（SED），仅对非零值进行建模，从而保留稀疏结构。SED在训练和推理中跳过零值，实现显著计算节省，同时保持甚至提升生成质量。在物理与生物学基准测试中，SED的表现与传统扩散模型或领域专用基线相当或更优；视觉实验则直观揭示了密集型扩散模型的局限性及SED的优势。
原文摘要 · Abstract (English)
Diffusion models (DMs) excel on dense continuous data, but are not designed for sparse continuous data. They do not model exact zeros that represent the deliberate absence of a signal. As a result, they erase sparsity patterns and perform unnecessary computation on mostly zero entries. With Sparsity-Exploiting Diffusion (SED), we model only non-zero values, preserving sparsity. SED delivers computational savings while maintaining or improving generation quality by skipping zeros during training and inference. Across physics and biology benchmarks, SED matches or surpasses conventional DMs and domain-specific baselines, while vision experiments provide intuitive insights into the limitations of dense DMs and the benefits of SED.
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。