隐私计算系统V2

1.任务调度：支持联邦训练任务的调度管理，并支持任务执行过程的可视化。 2.数据管理：支持数据导入、探索性数据分析、数据预处理和特征工程。 3.联邦训练：支持横向联邦训练与纵向联邦训练，具体算法包括：纵向联邦线性回归、纵向联邦逻辑回归、纵向联邦安全提升树、横向联邦逻辑回归、横向联邦安全提升树，以及面向场景的定制化算法设计。 4.模型评估：支持训练模型的量化评估，包括准确率、精确率（查准率）、召回率（查全率）、ROC曲线和K-S曲线等评估指标，同时支持训练模型的贡献评估。 5.模型管理：支持训练模型的查看、导出、部署、删除等操作。 6.在线预测：支持联邦在线预测，包括单例预测和批量预测，以及面向场景的定制化开发。 7.权限管理：支持创建不同操作权限的用户角色，实现角色与权限的动态配置。 8.系统日志：支持训练过程日志和用户操作日志。1. Task Scheduling: Supports scheduling and management of federated training tasks, as well as visualization of task execution processes. 2. Data Management: Supports data import, exploratory data analysis, data preprocessing, and feature engineering. 3. Federated Training: Supports both horizontal and vertical federated training. The supported specific algorithms include vertical federated linear regression, vertical federated logistic regression, vertical federated secure boosted trees, horizontal federated logistic regression, horizontal federated secure boosted trees, as well as scenario-oriented customized algorithm design. 4. Model Evaluation: Supports quantitative evaluation of trained models, including evaluation metrics such as accuracy, precision (also known as precision rate), recall (also known as recall rate), ROC curves, and K-S curves, and also supports contribution evaluation of trained models. 5. Model Management: Supports operations such as viewing, exporting, deploying, and deleting trained models. 6. Online Prediction: Supports federated online prediction, including single-instance prediction and batch prediction, as well as scenario-oriented customized development. 7. Permission Management: Supports creating user roles with different operation permissions, and realizes dynamic configuration of roles and permissions. 8. System Logs: Supports training process logs and user operation logs.

搜集汇总

背景与挑战

背景概述

该数据集详细描述了一个隐私计算系统V2，该系统支持联邦学习任务的全流程管理，包括数据导入与预处理、横向和纵向联邦训练算法、模型评估与部署，以及在线预测和权限控制。其特点在于提供多种联邦学习算法（如线性回归、逻辑回归和安全提升树），并支持定制化开发和可视化任务调度，适用于需要保护数据隐私的机器学习应用场景。