Synthetic Matrix Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/yhao-z/SVD-inv
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是为了测试可微奇异值分解方法而生成的合成矩阵集，其中包括了具有相近奇异值的矩阵。该数据集采用64位浮点精度生成矩阵，并包含了多种奇异值阈值化的工作流程。该数据集的规模为10x10的矩阵，其任务是对梯度计算和奇异值分析进行研究。
This dataset is a collection of synthetic matrices generated to test differentiable singular value decomposition (SVD) methods, which includes matrices with closely spaced singular values. All matrices in this dataset are generated using 64-bit floating-point precision, and the dataset incorporates multiple singular value thresholding workflows. Every matrix in this dataset has a size of 10×10, and this dataset is intended for research on gradient computation and singular value analysis.
提供机构：
Generated by the authors
