GilpinLab/skew40
收藏Hugging Face2025-07-15 更新2025-11-01 收录
下载链接：
https://hf-mirror.com/datasets/GilpinLab/skew40
下载链接
链接失效反馈官方服务：
资源简介：
skew40数据集是一个包含2万种混沌动力系统的合成数据集，用于训练Panda模型，该模型能够对未见的真实世界混沌系统进行零样本预测，并在交叉通道注意力头中表现出非线性共振模式。尽管仅在低维常微分方程上进行了训练，Panda模型自发地发展出了预测偏微分方程的能力。
The skew40 dataset is a synthetic dataset consisting of 20,000 chaotic dynamical systems used for training the Panda model, which is capable of zero-shot forecasting of unseen real-world chaotic systems and exhibits nonlinear resonance patterns in cross-channel attention heads. Despite being trained only on low-dimensional ordinary differential equations, the Panda model spontaneously develops the ability to predict partial differential equations without retraining.
提供机构：
GilpinLab
