Diversity-Based Hyperedge Sampling in GNN Learning | MICCAI MLMI 2024
#GNN #DeepLearning #MICCAI2024
"DHSampling: Diversity-based Hyperedge Sampling in GNN Learning with Application to Medical Image Classification", Workshop on MLMI, MICCAI 2024, [Jiameng Liu, Furkan Pala, Islem Rekik, and Dinggang Shen]
👉 Paper link: https://link.springer.com/chapter/10....
👉 arXiv link: TO-UPDATE
👉 Code link: https://github.com/basiralab/DHSampling
• Graph Neural Networks (GNNs) are powerful for handling heterogeneous clinical data, but face scalability issues due to substantial memory and computation demands.
• Existing graph-sampling strategies for efficient graph learning still need significant memory demands and may introduce bias due to unbalanced graph partitioning.
• Aims: (1) Hyperedge-based graph-partition for unbiased subgraph partition, (2) Diversity-based sampling for accelerate GNNs learning while keep comparable performance.