DataFunTalk

Jan 17, 2022 · Artificial Intelligence

Graph Attention Multi‑Layer Perceptron (GAMLP) and Node‑Dependent Local Smoothing (NDLS) for Scalable and Flexible Graph Neural Networks

This talk introduces the motivation, design, theoretical analysis, and extensive experimental results of Tencent Angel Graph's Graph Attention Multi‑Layer Perceptron (GAMLP) and Node‑Dependent Local Smoothing (NDLS), which address GNN scalability and flexibility by using node‑wise adaptive propagation, attention‑based feature fusion, and a lightweight training pipeline.