Microsoft Benchmark Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/Azure-Samples/azure-ml-federated-learning/blob/main/docs/concepts/benchmarking.md
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一组用于比较跨孤岛联邦学习与集中式学习在性能和可扩展性方面的基准数据集。这些数据集的大小涵盖了公司中大多数经典的机器学习用例，基准测试表明，通过跨孤岛联邦学习，模型的性能有所提升。数据集的规模分为小型（1.2 GB）、中型（12 GB）和大型（120 GB）。其任务是对跨孤岛联邦学习与集中式学习之间的二氧化碳排放量和成本进行比较分析。
This dataset is a benchmark suite for comparing the performance and scalability between cross-silo federated learning and centralized learning. The scales of these datasets cover most classic machine learning use cases in enterprises. Benchmark results demonstrate that cross-silo federated learning can improve model performance. The dataset is divided into three size tiers: small (1.2 GB), medium (12 GB), and large (120 GB). Its core task is to conduct a comparative analysis of carbon dioxide emissions and costs between cross-silo federated learning and centralized learning.
提供机构：
Microsoft
