IntraMix: Intra-Class Mixup Generation for Accurate Labels and Neighbors

arXiv:2405.00957v1 Announce Type: new
Abstract: Graph Neural Networks (GNNs) demonstrate excellent performance on graphs, with their core idea about aggregating neighborhood information and learning from labels. However, the prevailing challenges in most graph datasets are twofold of Insufficient High-Quality Labels and Lack of Neighborhoods, resulting in weak GNNs. Existing data augmentation methods designed to address these two issues often tackle only one. They may either require extensive training of generators, rely on overly simplistic strategies, or demand substantial prior knowledge, …

abstract arxiv augmentation challenges class core cs.ai cs.lg cs.si data datasets gnns graph graph neural networks graphs however information labels neighbors networks neural networks performance quality type