Graph Star Net for Generalized Multi-Task Learning

21 Jun 2019  ·  Lu Haonan, Seth H. Huang, Tian Ye, Guo Xiuyan ·

In this work, we present graph star net (GraphStar), a novel and unified graph neural net architecture which utilizes message-passing relay and attention mechanism for multiple prediction tasks - node classification, graph classification and link prediction. GraphStar addresses many earlier challenges facing graph neural nets and achieves non-local representation without increasing the model depth or bearing heavy computational costs. We also propose a new method to tackle topic-specific sentiment analysis based on node classification and text classification as graph classification. Our work shows that 'star nodes' can learn effective graph-data representation and improve on current methods for the three tasks. Specifically, for graph classification and link prediction, GraphStar outperforms the current state-of-the-art models by 2-5% on several key benchmarks.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Benchmark
Text Classification 20NEWS GraphStar Accuracy 86.9 # 8
Node Classification Citeseer GraphStar Accuracy 71.0 # 57
Link Prediction Citeseer (biased evaluation) GraphStar (double weight on positive examples) AUC 97.47 # 1
AP 97.93 # 1
Accuracy 97.7 # 1
Node Classification Cora GraphStar Accuracy 82.1% # 55
Link Prediction Cora (biased evaluation) GraphStar (double weight on positive examples) AUC 95.65 # 1
AP 96.15 # 1
Accuracy 95.9 # 1
Graph Classification D&D GraphStar Accuracy 79.60% # 15
Graph Classification ENZYMES GraphStar Accuracy 67.1% # 13
Sentiment Analysis IMDb GraphStar Accuracy 96.0 # 4
Sentiment Analysis MR GraphStar Accuracy 76.6 # 14
Graph Classification MUTAG GraphStar Accuracy 91.2% # 14
Text Classification Ohsumed GraphStar Accuracy 64.2 # 7
Node Classification PPI GraphStar F1 99.4 # 7
Graph Classification PROTEINS GraphStar Accuracy 77.90% # 19
Node Classification Pubmed GraphStar Accuracy 77.2% # 59
Link Prediction Pubmed (biased evaluation) GraphStar (double weight on positive examples) AUC 97.67 # 1
AP 98.64 # 1
Accuracy 98.16 # 1
Text Classification R52 GraphStar Accuracy 95.00 # 2
Text Classification R8 GraphStar Accuracy 97.4 # 12

Methods


No methods listed for this paper. Add relevant methods here