ARCH
收藏arXiv2025-09-30 收录
数据链接：
官方服务：
资源简介：
该数据集是一个用于评估重新制定的优化目标函数在学习等距变换中有效性的综合数据集。此外，该数据集还用于验证在潜在（子）流形上生成方法的有效性。其规模为样本量 n = 500，维度 d = 2。该数据集的任务是对图匹配损失和稳定性正则化进行消融研究。
This is a comprehensive benchmark dataset for evaluating the effectiveness of reformulated optimization objective functions in learning isometric transformations. Additionally, it is used to validate the effectiveness of generative methods on latent (sub)manifolds. It has a sample size of n=500 and a dimensionality of d=2. The task of this dataset is to conduct ablation studies on graph matching loss and stability regularization.
