Sequential Aggregation and Rematerialization: Distributed Full-batch Training of Graph Neural Networks on Large Graphs.
Hesham MostafaPublished in: MLSys (2022)
Keyphrases
- neural network
- training process
- graph structure
- graph representation
- graph theory
- graph matching
- directed graph
- graph theoretic
- weighted graph
- labeled graphs
- graph model
- graph mining
- graph construction
- distributed sensor networks
- bipartite graph
- series parallel
- graph data
- structural pattern recognition
- graph databases
- graph theoretical
- graph clustering
- graph structures
- graph isomorphism
- subgraph isomorphism
- pattern recognition
- training algorithm
- graph classification
- adjacency matrix
- distributed systems
- graph transformation
- feedforward neural networks
- graph properties
- multi layer perceptron
- spanning tree
- undirected graph
- graph search
- feed forward neural networks
- random graphs
- graph partitioning
- graph kernels
- maximum common subgraph
- connected graphs
- graph representations
- evolving graphs
- graph patterns
- artificial neural networks
- real world graphs
- maximum cardinality
- dynamic graph
- web graph
- finding the shortest path
- introduce a general framework
- inexact graph matching
- batch mode
- directed acyclic
- dense subgraphs
- reachability queries
- connected dominating set
- adjacency graph
- query graph
- maximal cliques
- neighborhood graph
- polynomial time complexity
- shortest path
- small world
- back propagation
- minimum spanning tree
- planar graphs
- multilayer perceptron
- directed acyclic graph
- community discovery