| 联邦个性化学习推荐系统研究 |
| |
| 引用本文： | 李康康,袁萌,林凡.联邦个性化学习推荐系统研究[J].现代教育技术,2022(2). |
| |
| 作者姓名： | 李康康 袁萌 林凡 |
| |
| 作者单位： | 江苏师范大学江苏省教育信息化工程技术研究中心;江苏师范大学智慧教育学院;厦门大学信息学院 |
| |
| 基金项目： | 国家自然科学基金青年项目“基于联邦学习的个性化学习推荐研究”(项目编号:62107022);教育部人文社会科学研究青年项目“教育人工智能隐私保护问题研究”(项目编号:21JYC880);江苏省研究生科研与实践创新计划项目“网络学习资源群体进化预警研究”(项目编号:KYCX21_2501)的阶段性研究成果。 |
| |
| 摘 要： | 当前,个性化学习推荐系统面临数据隐私保护、"冷启动"和法律约束等问题,而联邦学习作为近年来优秀的数据隐私保护机器学习技术解决方案,可有效解决这些问题。基于此,文章将联邦学习和个性化学习推荐相结合,设计了联邦个性化学习推荐系统。首先,文章分析了联邦个性化学习推荐系统的具体应用场景,包括横向联邦、纵向联邦、联邦强化三种。其次,文章分别针对这三种应用场景设计了相应的应用解决方案。最后,文章探讨了未来联邦个性化学习推荐系统面临的严峻挑战,以期帮助教育利益相关者在保护数据隐私的同时共享数据价值,最终实现更安全、更高质量的个性化学习推荐服务。
|
| 关 键 词： | 联邦学习 个性化学习推荐 数据隐私 联邦推荐算法 |
Research on Federated Personalized Learning Recommendation System |
| |
| LI Kang-kang,YUAN Meng,LIN Fan.Research on Federated Personalized Learning Recommendation System[J].Modern Educational Technology,2022(2). |
| |
| Authors: | LI Kang-kang YUAN Meng LIN Fan |
| |
| Institution: | (Educational Informatization Engineering Technology Research Center,Jiangsu Normal University,Xuzhou,Jiangsu,China 221116;School of Wisdom Education,Jiangsu Normal University,Xuzhou,Jiangsu,China 221116;School of Informatics,Xiamen University,Xiamen,Fujian,China 361000) |
| |
| Abstract: | At present,personalized learning recommendation systems are faced with problems including data privacy protection,“cold start”,and legal constraints,and as an excellent machine learning technology solution for data privacy protection in recent years,federated learning can effectively solve these problems.Based on this,this paper combined federated personalized learning recommendation and personalized learning recommendation,and designed a federated personalized learning recommendation system.Firstly,this paper analyzed the specific application scenarios of the federated personalized learning recommendation system,which included horizontal federation,vertical federation,and federation reinforcement.Secondly,corresponding application solutions for the three application scenarios were designed.Finally,this paper discussed the severe challenges faced by the future federated personalized learning recommendation system,in order to help education stakeholders share data value while protecting data privacy,and ultimately to achieve a safer and higher quality personalized learning recommendation service. |
| |
| Keywords: | federated learning personalized learning recommendation data privacy federated recommendation algorithm |
| 本文献已被 维普 等数据库收录！ |
|
