search

HKUST-KnowComp / DualMessagePassing

Source Code for AAAI 2022 paper "Graph Convolutional Networks with Dual Message Passing for Subgraph Isomorphism Counting and Matching"
MIT License
23 stars 5 forks source link

Test on other dataset #2

Open fuvty opened 1 year ago

fuvty commented 1 year ago

Thanks for the great work! I am also studying subgraph counting and would like to test your code with other query and target graphs. How should I convert the original data (say in networkx format) to the supported format of yours? In addition, I want to perform subgraph counting with undirected, homogeneous graphs, is that possible? Thanks a lot!

seanliu96 commented 1 year ago

Hi, thanks for your interest in our work. I think the simplest way is to refer to the official tutorial. Or you can convert to networkx format to gml format, which is further loaded by read_graphs_from_dir function.

fuvty commented 1 year ago

Thank you so much for your timely response! I think I am now able to generate the data in the graphs and pattern folders. Is there some way to generate other data needed for training? For example, the .pt files in the datasets folder. Additionally, if I set the graph and pattern to be bi-directional ( edge (0,1) exists iff edge (1,0) exists), will the model be able to perform counting on undirected graphs automatically?

seanliu96 commented 1 year ago

For the data preprocessing, I think the current code in NeuralSubgraphCounting can help you. Undirected graph matching has been evaluated in our paper such as Regular and Erdos-Renyi graphs.