未明确指出

本论文探讨了联邦学习在生存分析中的应用，特别是通过构建异构数据集来模拟真实世界的数据分布。研究团队来自米兰理工大学，他们提出了一种基于Dirichlet分布的数据分割算法，包括数量偏斜分割和标签偏斜分割，以创建具有不同异质性水平的联邦数据集。这些数据集用于评估联邦生存模型的鲁棒性，尤其是在处理非独立同分布(non-IID)数据时的表现。数据集的创建旨在解决生存分析中数据分布不均、不完整和保密性等问题，通过联邦学习提高模型训练的质量和隐私保护。

This paper explores the application of federated learning in survival analysis, particularly by constructing heterogeneous datasets to simulate real-world data distributions. A research team from Politecnico di Milano proposes a data partitioning algorithm based on Dirichlet distribution, including quantity-skewed partitioning and label-skewed partitioning to generate federated datasets with varying levels of heterogeneity. These datasets are utilized to evaluate the robustness of federated survival models, especially their performance when handling non-independent and identically distributed (non-IID) data. The creation of these datasets aims to address challenges including imbalanced data distributions, incomplete data and data confidentiality in survival analysis, while improving the quality of model training and enhancing privacy protection via federated learning.