用潜在空间扩散模型模拟物理系统，速度提升百倍且精度稳定。
Lost in Latent Space: An Empirical Study of Latent Diffusion Models for Physics Emulation
- 在自编码器潜空间生成，降低计算开销
- 压缩率高达1000倍时仍保持高精度
- 适合需要快速高精度模拟的科研与工程场景
扩散模型推理时计算成本高昂，限制了其作为快速物理模拟器的应用。在图像和视频生成中，通过在自编码器的潜空间而非像素空间生成，已有效缓解此问题。本文研究该策略是否适用于动力系统模拟及其代价。结果表明，潜空间模拟在高达1000倍的压缩率下仍表现出惊人的鲁棒性。此外，基于扩散的模拟器始终比非生成类方法更准确，并通过预测多样性来补偿不确定性。我们还总结了一系列关键的实用设计选择，涵盖从架构到优化器的训练要点。
原文摘要 · Abstract (English)
The steep computational cost of diffusion models at inference hinders their use as fast physics emulators. In the context of image and video generation, this computational drawback has been addressed by generating in the latent space of an autoencoder instead of the pixel space. In this work, we investigate whether a similar strategy can be effectively applied to the emulation of dynamical systems and at what cost. We find that the accuracy of latent-space emulation is surprisingly robust to a wide range of compression rates (up to 1000x). We also show that diffusion-based emulators are consistently more accurate than non-generative counterparts and compensate for uncertainty in their predictions with greater diversity. Finally, we cover practical design choices, spanning from architectures to optimizers, that we found critical to train latent-space emulators.
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。