Graph Convolution
Graph convolution is a technique used in graph neural networks (GNNs) to process data represented as graphs, aiming to learn informative representations of nodes and edges by aggregating information from their neighbors. Current research focuses on improving efficiency and scalability, particularly for large graphs, exploring novel architectures like windowed GNNs and post-training methods to reduce computational complexity, and addressing issues like over-smoothing through techniques such as corrected convolutions and adaptive kernel approaches. These advancements are significant for various applications, including recommendation systems, image processing, medical image analysis, and human activity recognition, where graph-structured data is prevalent.
Papers
On Classification Thresholds for Graph Attention with Edge Features
Kimon Fountoulakis, Dake He, Silvio Lattanzi, Bryan Perozzi, Anton Tsitsulin, Shenghao Yang
Analysis of Convolutions, Non-linearity and Depth in Graph Neural Networks using Neural Tangent Kernel
Mahalakshmi Sabanayagam, Pascal Esser, Debarghya Ghoshdastidar