Privacy-Preserving Distributed Link Predictions Among Peers in Online Classrooms Using Federated Learning
Project Overview
The document explores the use of Generative AI in education, focusing on Federated Learning (FL) for link prediction in Social Learning Networks (SLNs) within online classrooms. It highlights how FL addresses privacy concerns and overcomes the limitations of centralized learning by creating personalized models that capture both shared and classroom-specific interaction patterns. The findings reveal that these personalized FL models significantly outperform traditional centralized methods, enhancing model performance and providing valuable insights into the dynamics of student interactions across diverse educational settings. This innovative approach not only fosters improved learning outcomes but also respects student privacy, making it a promising avenue for future developments in AI-driven education.
Key Applications
Federated Learning for link prediction in Social Learning Networks
Context: Online classrooms, targeting students in various subjects including STEM and Humanities.
Implementation: Federated Learning framework that integrates model personalization techniques.
Outcomes: Improved prediction accuracy and fairness in link prediction compared to centralized models.
Challenges: Privacy concerns regarding data sharing and the need for model personalization to account for diverse interaction patterns.
Implementation Barriers
Privacy Concern
Privacy regulations restrict the aggregation and sharing of student data across classrooms.
Proposed Solutions: Utilizing Federated Learning to enable model training without raw data centralization.
Data Diversity
Classrooms exhibit unique interaction patterns that can lead to suboptimal model performance with centralized approaches.
Proposed Solutions: Implementing model personalization techniques to adapt models to individual classroom characteristics.
Project Team
Anurata Prabha Hridi
Researcher
Muntasir Hoq
Researcher
Zhikai Gao
Researcher
Collin Lynch
Researcher
Rajeev Sahay
Researcher
Seyyedali Hosseinalipour
Researcher
Bita Akram
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Anurata Prabha Hridi, Muntasir Hoq, Zhikai Gao, Collin Lynch, Rajeev Sahay, Seyyedali Hosseinalipour, Bita Akram
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai