ConDiff
收藏资源简介:
ConDiff是一个专为科学机器学习设计的新型数据集,专注于具有变化系数的扩散方程,这是参数偏微分方程(PDEs)众多应用中的一个基础问题。该数据集的主要创新点在于考虑了高对比度的不连续系数,这些系数函数从选定的分布中采样而来。ConDiff不仅具有重大的学术价值,而且还是描述各种环境和工业问题的基石。数据集包含一系列多样化的扩散方程,其系数覆盖了广泛的对比度水平和异质性,并具有可测量的复杂度指标,以便于不同系数函数之间的清晰比较。ConDiff的应用领域包括复合材料建模、热传递、地球物理问题和流体流动建模,旨在解决这些领域中由于参数PDEs带来的复杂性问题。
ConDiff is a novel dataset specifically designed for scientific machine learning, focusing on diffusion equations with variable coefficients, which is a fundamental problem in numerous applications of parametric partial differential equations (PDEs). The core innovation of this dataset lies in its consideration of high-contrast discontinuous coefficients, where the coefficient functions are sampled from selected distributions. ConDiff not only holds significant academic value but also serves as a cornerstone for describing various environmental and industrial problems. The dataset includes a diverse set of diffusion equations, whose coefficients cover a wide range of contrast levels and heterogeneity, and is equipped with measurable complexity metrics to enable clear comparisons between different coefficient functions. The application scope of ConDiff covers composite material modeling, heat transfer, geophysical problems, and fluid flow modeling, aiming to address the complexity issues caused by parametric PDEs in these fields.

- 1ConDiff: A Challenging Dataset for Neural Solvers of Partial Differential Equations斯科尔科沃科技学院 · 2024年



