Training Grids

该数据集包含了1000个随机生成的网格，用于训练MG-GNN模型，网格大小在800到1000个节点之间。这些网格被创建为凸多边形，子域则是通过Lloyd聚类方法生成。该数据集用于评估模型的性能，其生成方式能够扩展到更大的测试网格，节点数量从800到60,000个自由度不等。该数据集的规模为1000个网格，节点数量从800到1000个。所承担的任务是利用图神经网络优化双层域分解方法中的参数。

This dataset contains 1,000 randomly generated grids for training the MG-GNN model, with each grid ranging from 800 to 1,000 nodes. These grids are constructed as convex polygons, and their subdomains are generated via the Lloyd clustering method. This dataset is used for model performance evaluation, and its generation approach can be scaled to larger test grids where the number of nodes (or degrees of freedom) varies from 800 to 60,000. The dataset consists of 1,000 grids with node counts spanning 800 to 1,000. The targeted task is to optimize the parameters in the two-level domain decomposition method using graph neural networks.