This is the implementation of paper "GFinder: Approximate attributed subgraph matching".
History: At the begining, the algorithm was named FastDynamic and I submitted the paper to Ijcai, but it was rejected by Ijcai. Then, I renamed the algorithm to GLooker, and submitted it to CIKM, but it was rejected again. Finally, I renamed the algorithm to GFinder, and the paper was accepted.
If you have anyquestion about the algorithm, please open an issue.
Updated 2020-Dec-12. Another version of G-Finder which can support edge type similarity can be found at https://github.com/lihuiliullh/GFinder-Proj It contains some docx which tell details of how to run G-Finder. Its test data can also be used here.
This Project (G-Looker) is implemented on visual studio C++ 2015, windows 10 64 bits. Some sample data files are stored in direcotry "./testdata"
Usage:
You can use visual studio to open the project, and run the project.
--or
If you want to run G-Looker, use the command like this:
FastDynamic.exe data_file_name query_file_name top-k node-miss-cost intermediate-vertex-cost edge-miss-cost hop-number
For example:
.\FastDynamic.exe C:\flickr.txt.format C:\query_noise_7.txt.format 10 30 10 10 0
data_file_name: The data file query_file_name: The query file top-k: how many results you would like to find. node-miss-cost: missing node cost used in loss function (see our paper for more details) intermediate-vertex-cost: intermediate vertex cost used in loss function (see our paper for more details) edge-miss-cost: missing edge cost used in loss function (see our paper for more details) hop-number: if hop-number == 0, means 2 hop. If hop-number >=1 means 1 hop.
For example:
.\FastDynamic.exe C:\flickr.txt.format C:\query_noise_7.txt.format 10 30 10 10 0 (use two-hop distance)
.\FastDynamic.exe C:\flickr.txt.format C:\query_noise_7.txt.format 10 30 10 10 1 (use one-hop distance)
.\FastDynamic.exe C:\flickr.txt.format C:\query_noise_7.txt.format 10 30 10 10 2 (use one-hop distance)