Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Neural Subgraph Isomorphism Counting

About

In this paper, we study a new graph learning problem: learning to count subgraph isomorphisms. Different from other traditional graph learning problems such as node classification and link prediction, subgraph isomorphism counting is NP-complete and requires more global inference to oversee the whole graph. To make it scalable for large-scale graphs and patterns, we propose a learning framework which augments different representation learning architectures and iteratively attends pattern and target data graphs to memorize subgraph isomorphisms for the global counting. We develop both small graphs (<= 1,024 subgraph isomorphisms in each) and large graphs (<= 4,096 subgraph isomorphisms in each) sets to evaluate different models. A mutagenic compound dataset, MUTAG, is also used to evaluate neural models and demonstrate the success of transfer learning. While the learning based approach is inexact, we are able to generalize to count large patterns and data graphs in linear time compared to the exponential time of the original NP-complete problem. Experimental results show that learning based subgraph isomorphism counting can speed up the traditional algorithm, VF2, 10-1,000 times with acceptable errors. Domain adaptation based on fine-tuning also shows the usefulness of our approach in real-world applications.

Xin Liu, Haojie Pan, Mutian He, Yangqiu Song, Xin Jiang, Lifeng Shang• 2019

Related benchmarks

TaskDatasetResultRank
Structural Edit Distance (SED)AMAZON
Running Time (s)21
12
Structural Edit Distance (SED)Citeseer
Running Time (s)24.46
12
Structural Edit Distance (SED)AIDS
Latency (s)4
12
Structural Edit Distance (SED)Pubmed
Running Time (s)35.05
12
Structural Edit Distance (SED)PROTEIN
Running Time (s)21
12
Structural Edit Distance (SED)Cora-ML
Running Time (s)70.59
12
k-NN Ranking (SED)Pubmed
Kendall's tau0.89
10
k-NN Ranking (SED)PROTEIN
Kendall's Tau0.74
10
k-NN Ranking (SED)Citeseer
Kendall's Tau0.88
10
k-NN Ranking (SED)Cora-ML
Kendall's Tau0.88
10
Showing 10 of 17 rows

Other info

Follow for update