The paper proposes CATGNN, a distributed training system for graph neural networks (GNNs) that addresses the limitations of existing approaches. Key highlights:
CATGNN takes a stream of edges as input for graph partitioning, instead of loading the entire graph into memory, enabling training on large-scale graphs under limited computational resources.
CATGNN adopts a novel streaming partitioning algorithm called SPRING that leverages the 'richest neighbor' information to improve partitioning quality and reduce the replication factor.
CATGNN uses model averaging for model synchronization across workers, which avoids the issues with gradient averaging and reduces communication overhead.
CATGNN is highly flexible and extensible, allowing users to integrate custom streaming partitioning algorithms and GNN models.
Experiments show CATGNN can handle the largest publicly available dataset with limited memory, which would have been infeasible with existing approaches. SPRING also outperforms state-of-the-art streaming partitioning algorithms by reducing the replication factor by 50% on average.
To Another Language
from source content
arxiv.org
Viktige innsikter hentet fra
by Xin Huang,We... klokken arxiv.org 04-04-2024
https://arxiv.org/pdf/2404.02300.pdfDypere Spørsmål