English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Graph Convolutional Networks (GCNs) are extensively utilized for deeplearning on graphs. The large data sizes of graphs and their vertex featuresmake scalable training algorithms and distributed memory systems necessary.Since the convolution operation on graphs induces irregular memory accesspatterns, designing a memory- and communication-efficient parallel algorithmfor GCN training poses unique challenges. We propose a highly parallel trainingalgorithm that scales to large processor counts. In our solution, the largeadjacency and vertex-feature matrices are partitioned among processors. Weexploit the vertex-partitioning of the graph to use non-blocking point-to-pointcommunication operations between processors for better scalability. To furtherminimize the parallelization overheads, we introduce a sparse matrixpartitioning scheme based on a hypergraph partitioning model for full-batchtraining. We also propose a novel stochastic hypergraph model to encode theexpected communication volume in mini-batch training. We show the merits of thehypergraph model, previously unexplored for GCN training, over the standardgraph partitioning model which does not accurately encode the communicationcosts. Experiments performed on real-world graph datasets demonstrate that theproposed algorithms achieve considerable speedups over alternative solutions.The optimizations achieved on communication costs become even more pronouncedat high scalability with many processors. The performance benefits arepreserved in deeper GCNs having more layers as well as on billion-scale graphs.

Scalable Graph Convolutional Network Training on Distributed-Memory Systems