Update: All code has been revised to include random edge scalability, massive knowledge graph construction, and parallel computing.

POMA

This repository contains c++ codes and datasets for the paper:

Efficient Maximal Motif-Clique Enumeration over Large Heterogeneous Information Networks

Environment

The codes of POMA are implemented and tested under the following development environment:

• Hardware : Intel(R) Xeon(R) Gold 6226R 2.90GHz CPU and 512GB of memory.

• Operation System : Ubuntu 20.04.4 LTS (GNU/Linux 5.13.0-40-generic x86_64)

Datasets

We use five real HINs: Instacart, WordNet, DBLP,DBpedia, and Freebase. The dataset file in this Google driven link. The detailed statistics and original sources of these datasets are in our paper.

How to Run the Codes

A. Code Compilation

You can just use the following commands:

make clean
cmake .
make

B. Data format

The data format is:

1. the number of vertices n, the number of vertices m

2. the edges

3. the vertex type of the all vertices.

n,m
id_1, id_2
id_3, id_4
....
type_1, type_2, type_3, ..., type_n

C. Run Code

You can run the shell scripts like xxx.sh under META/scripts and POMA/scripts for the five datasets.

You may need to modify the paths in the shell scripts to meet the corresponding folders.

Such as you can just run bash run_dblp.sh to run our POMA algorithm for the dblp dataset.

Also, you can use the following parameters to obtain the different version of POMA.

-ORDER 0	not use order POMA-O
-InitVF3 1	not use the fast candidate set algorithm POMA-C
-PIVOT 0	not use motif-pivot POMA-P
--NotReduce 1	not use graph reduce POMA-R

For the degree and degeneracy order, you can:

-ORDER 1	degree order
-ORDER 2	degeneracy order

For the Scalability experiment, you can just run bash run_dblp_sca.sh to obtain the result for the dblp dataset.

Similarly, you can modify the dataset name to obtain the results for other datasets.

C. Data Download

You can download the datasets from the following Google driven link:

https://drive.google.com/file/d/1RNVe5LcwN_j5S5XoJJgVNcHMuBaezlVW/view?usp=sharing