arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.22122stat.MLcs.LG
收藏

用字符串方法探究扩散模型的几何结构

Probing the Geometry of Diffusion Models with the String Method

Capital Fund Management(资本基金管理公司);ML Lab, Capital Fund Management(机器学习实验室,资本基金管理公司);Courant Institute of Mathematical Sciences, New York University(Courant数学科学研究所,纽约大学)

Elio Moreau, Florentin Coeurdoux, Grégoire Ferre, Eric Vanden-Eijnden

AI总结:

本文提出基于字符串方法的框架,用于探究扩散模型的几何结构,通过不同模式插值揭示模式、障碍物和连接性。

摘要:

理解学习到的分布的几何结构对于改进和解释扩散模型至关重要,但系统地探索其景观的工具仍然有限。标准的潜在空间插值无法尊重学习到的分布的结构,往往穿越低密度区域。我们引入了一种基于字符串方法的框架,通过在学习到的分数函数下演化曲线来计算样本之间的连续路径。在不重新训练的情况下,我们的方法在三种模式之间进行插值:纯生成运输,产生连续的样本路径;梯度主导的动力学,恢复最小能量路径(MEPs);有限温度的字符串动力学,计算主曲线——自洽的路径,平衡能量和熵。我们证明了在实践中模式的选择是有影响的。对于图像扩散模型,MEPs包含高似然但不现实的''卡通''图像,证实了先前的观察,即似然最大值似乎不现实;主曲线反而产生现实的变形序列,尽管似然较低。对于蛋白质结构预测,我们的方法直接从训练在静态结构上的模型中计算介稳构象之间的过渡路径,产生具有物理上合理的中间体的路径。这些结果确立了字符串方法作为探究扩散模型模式结构的原理性工具——识别模式、表征障碍物并映射复杂学习分布中的连接性。

英文摘要:

Understanding the geometry of learned distributions is fundamental to improving and interpreting diffusion models, yet systematic tools for exploring their landscape remain limited. Standard latent-space interpolations fail to respect the structure of the learned distribution, often traversing low-density regions. We introduce a framework based on the string method that computes continuous paths between samples by evolving curves under the learned score function. Operating on pretrained models without retraining, our approach interpolates between three regimes: pure generative transport, which yields continuous sample paths; gradient-dominated dynamics, which recover minimum energy paths (MEPs); and finite-temperature string dynamics, which compute principal curves -- self-consistent paths that balance energy and entropy. We demonstrate that the choice of regime matters in practice. For image diffusion models, MEPs contain high-likelihood but unrealistic ''cartoon'' images, confirming prior observations that likelihood maxima appear unrealistic; principal curves instead yield realistic morphing sequences despite lower likelihood. For protein structure prediction, our method computes transition pathways between metastable conformers directly from models trained on static structures, yielding paths with physically plausible intermediates. Together, these results establish the string method as a principled tool for probing the modal structure of diffusion models -- identifying modes, characterizing barriers, and mapping connectivity in complex learned distributions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑