Adaptive Message Quantization and Parallelization for Distributed Full-graph GNN Training.
Borui WanJuntao ZhaoChuan WuPublished in: MLSys (2023)
Keyphrases
- distributed environment
- distributed systems
- multi agent
- graph theory
- graph representation
- training set
- random walk
- parallel execution
- training process
- graph structure
- supervised learning
- peer to peer
- training samples
- structured data
- cooperative
- parallel processing
- bipartite graph
- computing environments
- test set
- lightweight
- training data
- directed graph
- directed acyclic graph
- graph mining
- graph databases
- spanning tree
- communication channels
- training phase
- commit protocols
- subband image coding