ProFed
收藏资源简介:
ProFed是一个针对 proximity-based 非独立同分布(non-IID)联邦学习的新型基准测试。该数据集由博洛尼亚大学Cesena创建,旨在通过模拟不同区域的数据分布偏差,为研究人员提供标准化的框架,以更有效和一致地评估联邦学习算法。ProFed利用了知名的计算机视觉数据集,如MNIST、FashionMNIST、CIFAR-10和CIFAR-100,并采用了文献中的数据划分方法,如基于Dirichlet分布的划分。通过允许研究人员控制数据偏斜程度,该方法可以进行细致的实验和分析。
ProFed is a novel benchmark for proximity-based non-independent and identically distributed (non-IID) federated learning. Developed by the University of Bologna's Cesena Campus, this dataset aims to provide researchers with a standardized framework to evaluate federated learning algorithms more efficiently and consistently by simulating data distribution biases across distinct regions. ProFed leverages well-established computer vision datasets including MNIST, FashionMNIST, CIFAR-10, and CIFAR-100, and adopts data partitioning methods from prior literature, such as Dirichlet distribution-based partitioning. By permitting researchers to control the degree of data skew, this framework supports detailed experimental investigations and comprehensive analysis.
ProFed数据集概述
基本信息
- 数据集名称:ProFed: A Benchmark for Proximity-based Federated Learning
数据集简介
- ProFed是一个基于邻近性的联邦学习基准测试数据集。
应用领域
- 联邦学习
- 邻近性计算
- 机器学习基准测试
特点
- 专注于邻近性计算在联邦学习中的应用
- 提供基准测试功能

- 1ProFed: a Benchmark for Proximity-based non-IID Federated Learning博洛尼亚大学Cesena · 2025年



