In this paper, we present Deep Graph Kernels (DGK), a unified framework to learn latent representations of sub-structures for graphs, inspired by latest advancements in language modeling and deep learning. Our framework leverages the dependency information between sub-structures by learning their latent representations. We demonstrate instances of our framework on three popular graph kernels, namely Graphlet kernels, Weisfeiler-Lehman subtree kernels, and Shortest-Path graph kernels. Our experiments on several benchmark datasets show that Deep Graph Kernels achieve significant improvements in classification accuracy over state-of-the-art graph kernels.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Graph Classification COLLAB DGK Accuracy 73.09% # 25
Graph Classification D&D DGK Accuracy 73.50% # 41
Graph Classification ENZYMES DGK Accuracy 53.43% # 30
Graph Classification IMDb-B DGK Accuracy 66.96% # 38
Graph Classification IMDb-M DGK Accuracy 44.55% # 32
Graph Classification MUTAG DGK Accuracy 87.44% # 44
Graph Classification PROTEINS DGK Accuracy 75.68% # 50
Graph Classification RE-M12K DGK Accuracy 32.22% # 6
Graph Classification RE-M5K DGK Accuracy 41.27% # 8

Results from Other Papers

Task Dataset Model Metric Name Metric Value Rank Source Paper Compare
Malware Clustering Android Malware Dataset Deep WL kernel ARI 50.41 # 2
Malware Detection Android Malware Dataset Deep WL kernel Accuracy 98.16 # 2


No methods listed for this paper. Add relevant methods here