IRM Experiments
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/facebookresearch/InvariantRiskMinimization
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了来自三个训练环境的观察变异，旨在评估不变风险最小化（IRM）算法的泛化能力。具体来说，数据集涵盖了基于打乱和未打乱的观察、完全观察和部分观察的图形，以及不同类型的噪声。规模上，该数据集从三个训练环境中抽取了1000个样本，任务是对多个环境下的泛化能力进行评估。
This dataset includes observational variations across three training environments, designed to evaluate the generalization capability of the Invariant Risk Minimization (IRM) algorithm. Specifically, it covers observations with and without shuffling, fully observed and partially observed graphs, as well as various types of noise. In terms of scale, 1000 samples are collected from the three training environments, and the dataset is intended for assessing generalization capabilities across multiple environments.
提供机构：
Authors of the paper
