DynaBench
收藏资源简介:
DynaBench是由德国维尔茨堡大学创建的一个新型模拟基准数据集,专注于从稀疏分布的数据中学习动力系统。该数据集包含7000个模拟,涵盖六种不同的偏微分方程系统,用于评估机器学习模型在预测动力系统演化方面的能力。数据集通过模拟生成,无需预先了解方程,更贴近真实世界的数据获取方式。DynaBench旨在为机器学习社区提供一个易于使用的工具,以评估在只有非结构化低分辨率观测数据可用的情况下的模型性能。
DynaBench is a novel simulated benchmark dataset developed by the University of Würzburg, Germany, focusing on learning dynamical systems from sparsely distributed data. This dataset includes 7000 simulations covering six distinct partial differential equation (PDE) systems, and is designed to evaluate the capability of machine learning models in predicting the evolution of dynamical systems. Generated via numerical simulations, the dataset requires no prior knowledge of the underlying equations, making it more consistent with real-world data acquisition practices. DynaBench aims to provide the machine learning community with an easy-to-use tool for assessing model performance when only unstructured low-resolution observational data is available.

- 1DynaBench: A benchmark dataset for learning dynamical systems from low-resolution data维尔茨堡大学 · 2023年



