Highlights from the CNCC 2021 Knowledge Graph and Graph Machine Learning Forum

Recently, the CNCC 2021 "Knowledge as Intent, Graphs as Form – Knowledge Graph Reasoning Based on Graph Machine Learning" technical forum was held, co‑organized by Beike Zhaofang and scholars from the University of Science and Technology of China, Texas A&M University, Michigan State University, Tencent and others. The event focused on developing deep‑learning techniques better suited for knowledge graphs, improving modeling capabilities, and expanding industrial use cases.

The forum was chaired by Dr. Ye Jiepeng, Vice President and Chief Scientist of Beike Zhaofang, together with Prof. Wang Jie from USTC. Speakers included Prof. Ji Shuiwang (Texas A&M), Associate Prof. Tang Jiliang (Michigan State), Dr. Feng Yang (Beike AI Center), and Mr. Ma Jianqiang (Tencent), who provided in‑depth analyses of the latest research and industry challenges.

Speaker Tang Jiliang presented "Understanding and Designing Graph Neural Networks from an Optimization Perspective," proposing that many GNN models can be unified as graph denoising optimization problems and suggesting new design directions based on this view.

Speaker Ji Shuiwang delivered "Graph Computing for Quantum Chemistry and Physics," introducing the SphereNet architecture for handling 3‑D molecular graphs and the DIG and MoleculeX code frameworks that address distance, angle, and torsion information in quantum‑system modeling.

Speaker Wang Jie shared "Reasoning Techniques on Knowledge Graphs – From Simple to Complex," reviewing knowledge‑graph fundamentals, rule‑based and representation‑learning based reasoning, and showcasing his lab’s recent advances in both simple and complex reasoning tasks.

Speaker Feng Yang discussed "Applications of Graph Technologies in the New Housing Industry," describing how Beike employs relational and knowledge graphs for property recommendation, intelligent assistants, risk control, and AI‑driven training platforms.

Speaker Ma Jianqiang concluded with "Industrial Practices of Domain Knowledge Graphs at Tencent," outlining low‑risk, cost‑effective graph construction driven by specific video‑search scenarios, multi‑strategy entity alignment, multimodal linking, and graph‑based recommendation.

The forum attracted over a hundred live viewers, fostering lively interaction between academia and industry and highlighting the growing importance of graph‑based AI techniques across diverse domains.