GraphVite: A High-Performance CPU-GPU Hybrid System for Node Embedding

2 Mar 2019Zhaocheng ZhuShizhen XuMeng QuJian Tang

Learning continuous representations of nodes is attracting growing interest in both academia and industry recently, due to their simplicity and effectiveness in a variety of applications. Most of existing node embedding algorithms and systems are capable of processing networks with hundreds of thousands or a few millions of nodes... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Link Prediction FB15k SimplE MR 74 # 2
Link Prediction FB15k SimplE MRR 0.779 # 4
Link Prediction FB15k SimplE [email protected] 0.876 # 4
Link Prediction FB15k SimplE [email protected] 0.818 # 3
Link Prediction FB15k SimplE [email protected] 0.721 # 3
Link Prediction FB15k SimplE training time (s) 2105 # 1
Link Prediction FB15k-237 RotatE MRR 0.314 # 6
Link Prediction FB15k-237 RotatE [email protected] 0.511 # 5
Link Prediction FB15k-237 RotatE [email protected] 0.347 # 5
Link Prediction FB15k-237 RotatE [email protected] 0.217 # 5
Link Prediction FB15k-237 RotatE MR 176 # 1
Link Prediction FB15k-237 RotatE training time (s) 857 # 1
Node Classification Flickr LINE [email protected]% 63.05 # 1
Node Classification Flickr LINE [email protected]% 61.77 # 1
Node Classification Flickr LINE training time (s) 212 # 1
Link Prediction WN18 SimplE MRR 0.948 # 5
Link Prediction WN18 SimplE [email protected] 0.954 # 4
Link Prediction WN18 SimplE [email protected] 0.950 # 3
Link Prediction WN18 SimplE [email protected] 0.944 # 4
Link Prediction WN18 SimplE MR 412 # 3
Link Prediction WN18 SimplE training time (s) 1042 # 1
Link Prediction WN18RR RotatE MRR 0.490 # 2
Link Prediction WN18RR RotatE [email protected] 0.589 # 2
Link Prediction WN18RR RotatE [email protected] 0.508 # 1
Link Prediction WN18RR RotatE [email protected] 0.439 # 4
Link Prediction WN18RR RotatE MR 1845 # 1
Link Prediction WN18RR RotatE training time (s) 828 # 1
Node Classification YouTube LINE runtime (s) 70.09 # 1
Node Classification YouTube LINE [email protected]% 40.61 # 1
Node Classification YouTube LINE [email protected]% 33.69 # 1