Autobalanced Multitask Node Embedding Framework for Intelligent Education

Area of Science:

• Computer Science
• Educational Technology
• Artificial Intelligence

Background:

• Online education platforms increasingly use intelligent services for learning enhancement.
• Node embedding is crucial for these services but faces challenges with heterogeneous, evolving, and imbalanced educational graphs.
• Existing methods struggle with the unique properties of educational interaction graphs.

Purpose of the Study:

• To propose an autobalanced multitask node embedding model (MNE) for intelligent education.
• To address the challenges of heterogeneity, evolution, and lopsidedness in educational graphs.
• To improve learning efficiency and effectiveness through advanced node embedding.

Main Methods:

• Developed MNE, an autobalanced multitask node embedding model.
• Implemented two self-supervised learning tasks: edge-specific reconstruction and evolutive weight regression.
• Integrated uncertainty quantification for task- and node-level weight estimation and subtask autobalancing.

Main Results:

• MNE outperforms state-of-the-art graph embedding methods on real-world datasets.
• Demonstrated the effectiveness of the multitask framework and autobalancing mechanism.
• Successfully applied MNE to practical tasks in intelligent education.

Conclusions:

• The proposed MNE model effectively handles complex educational graph properties.
• MNE offers a valid multitask framework with a robust subtask balancing mechanism.
• This approach enhances intelligent services in online education by improving node embedding.