Attribute-device

该数据集是从属性隔离数据集中衍生出来的，通过将每个状态分割成六个子集，并对其进行抽样，以创建满足偏见约束的数据库。观察到的子集之间的偏见模式，继承了属性隔离设置中相应父状态数据集的偏见模式。该数据集适用于跨设备场景，其任务是评估联邦学习中的公平性。This dataset is derived from the attribute isolation dataset. It is constructed by splitting each state into six subsets and performing sampling on them to create a database that satisfies bias constraints. The observed bias patterns among the subsets inherit those from the corresponding parent-state dataset under the attribute isolation setting. This dataset is applicable to cross-device scenarios, and its task is to evaluate fairness in federated learning.