Skip to main content Skip to navigation

GNN 101: Visual Learning of Graph Neural Networks in Your Web Browser

Project Overview

The document discusses the innovative use of generative AI in education, particularly through the implementation of GNN101, an interactive educational visualization tool aimed at enhancing the learning experience for Graph Neural Networks (GNNs). By combining mathematical formulas with visual representations, GNN101 offers a user-friendly interface that supports comprehension through hierarchical levels of detail and complementary views, including node-link and matrix formats. The tool facilitates interactive math-visualization linking, which has proven effective in promoting a deeper understanding of complex GNN concepts. Deployed in three university courses, GNN101 has shown positive outcomes in usability and effectiveness, benefiting both students and instructors by aiding in the grasp of intricate subject matter. Overall, the integration of generative AI in educational tools like GNN101 highlights its potential to transform learning processes and improve educational outcomes in specialized fields.

Key Applications

GNN101

Context: Educational tool for learning about Graph Neural Networks; target audience includes computer science students, teaching assistants, and instructors.

Implementation: GNN101 was developed based on user feedback and expert collaboration, deployed in courses to facilitate understanding of GNNs.

Outcomes: Improved understanding of message passing, GNN variants, and GNN tasks among students and TAs; positive feedback on usability and engagement.

Challenges: Initial confusion in using visualizations, need for clear guidance and onboarding for new users.

Implementation Barriers

User Experience Barrier

Students found some visual representations confusing, particularly with adjacency matrices. Onboarding tutorials and step-by-step animations were implemented to guide users and reduce cognitive load.

Technical Barrier

Difficulty in dynamically modifying graph data and visualizations in real-time. Future iterations could introduce dynamic graph modification features.

Project Team

Yilin Lu

Researcher

Chongwei Chen

Researcher

Yuxin Chen

Researcher

Kexin Huang

Researcher

Marinka Zitnik

Researcher

Qianwen Wang

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Yilin Lu, Chongwei Chen, Yuxin Chen, Kexin Huang, Marinka Zitnik, Qianwen Wang

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

Let us know you agree to cookies