[2010.05337] DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphsopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

Graph neural networks (GNN) have shown great success in learning from graph-structured data. They are widely used in various applications, such as recommendation, fraud detection, and search. In these domains, the graphs are typically large, containing hundreds of millions of nodes and several billions of edges. To tackle this challenge, we develop DistDGL, a system for training GNNs in a mini-batch fashion on a cluster of machines. DistDGL is based on the Deep Graph Library (DGL), a popular GNN development framework. DistDGL distributes the graph and its associated data (initial features and embeddings) across the machines and uses this distribution to derive a computational decomposition by following an owner-compute rule. DistDGL follows a synchronous training approach and allows ego-networks forming the mini-batches to include non-local nodes. To minimize the overheads associated with distributed computations, DistDGL uses a high-quality and light-weight min-cut graph partitioning

3 mentions: @smolix@tristanzajonc
Date: 2020/10/14 12:53

Referring Tweets

@smolix Linear scaling for distributed training on graphs with 1 Billion nodes. Check out DistDGL t.co/cYoWAPAPHF t.co/4VOJChJEfL

Related Entries

Read more Using imitation and reinforcement learning to tackle long-horizon robotic tasks
0 users, 13 mentions 2019/11/13 18:51
Read more AlphaFold: Using AI for scientific discovery | DeepMind
4 users, 51 mentions 2020/01/17 08:21
Read more [2003.02989] TensorFlow Quantum: A Software Framework for Quantum Machine Learningcontact arXivarXiv...
1 users, 29 mentions 2020/03/09 06:52
Read more Exploring Bayesian OptimizationExploring Bayesian Optimization
0 users, 11 mentions 2020/05/05 23:21
Read more What is MLOps? | NVIDIA Blog Artificial Intelligence...
0 users, 30 mentions 2020/09/03 15:00