arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.08733cs.LG
收藏

常微分方程的基础推理模型

Foundation Inference Models for Ordinary Differential Equations

University of Cambridge(剑桥大学);ETH Zurich(苏黎世联邦理工学院)

Maximilian Mauel, Johannes R. Hübers, David Berghaus, Patrick Seifner, Ramses J. Sanchez

AI总结:

提出FIM-ODE,一种预训练的基础推理模型,通过单次前向传播从含噪轨迹直接预测向量场,实现零样本性能匹配并超越ODEFormer,微调后优于现代神经和GP基线。

摘要:

常微分方程(ODE)是科学建模的核心,但从含噪轨迹中推断其向量场仍然具有挑战性。当前的方法,如符号回归、高斯过程(GP)回归和神经常微分方程,通常需要复杂的训练流程和大量的机器学习专业知识,或者严重依赖于系统特定的先验知识。我们提出FIM-ODE,一种预训练的基础推理模型,通过单次前向传播直接从含噪轨迹数据预测向量场,从而摊销低维ODE推理。我们在具有低次多项式向量场的ODE先验分布上预训练FIM-ODE,并用神经算子表示目标场。FIM-ODE实现了强大的零样本性能,在多种设置下匹配并常常优于最近的预训练符号基线ODEFormer,尽管使用了更简单的预训练先验分布。预训练还为微调提供了强大的初始化,实现了快速且稳定的适应,在不需要机器学习专业知识的情况下优于现代神经和GP基线。

英文摘要:

Ordinary differential equations (ODEs) are central to scientific modelling, but inferring their vector fields from noisy trajectories remains challenging. Current approaches such as symbolic regression, Gaussian process (GP) regression, and Neural ODEs often require complex training pipelines and substantial machine learning expertise, or they depend strongly on system-specific prior knowledge. We propose FIM-ODE, a pretrained Foundation Inference Model that amortises low-dimensional ODE inference by predicting the vector field directly from noisy trajectory data in a single forward pass. We pretrain FIM-ODE on a prior distribution over ODEs with low-degree polynomial vector fields and represent the target field with neural operators. FIM-ODE achieves strong zero-shot performance, matching and often improving upon ODEFormer, a recent pretrained symbolic baseline, across a range of regimes despite using a simpler pretraining prior distribution. Pretraining also provides a strong initialisation for finetuning, enabling fast and stable adaptation that outperforms modern neural and GP baselines without requiring machine learning expertise.

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。

↑