FastPat-KG

OVERVIEW

• FastPat is a framework to solve kCP problem, which extracts core-based top-k frequent patterns in knowledge graphs.
• FastPat+ is a framework to solve kCPD problem, which extracts core-based top-k frequent patterns with diversity in knowledge graphs.

REQUIREMENTS

Java JRE v1.8.0 or later

USAGE

Settings

The proposed optimizations of each algorithm (e.g., kCP-A, FPD-A) could be set in the configuration file Settings.java.

Instructions to compile/run

- Uncompress FastPat-KG by any compression tool
- Build Java binaries using the "build" script file
- Run FastPat/FastPat+ using "topkFreq/topkFreqDiv" script

Execution examples

1. kCP problem

• Extract Top-10 frequent patterns in "Oscar" knowledge graph, with k = 10, core pattern is OscarTest: "./topkFreq.sh -f Oscar.lg -cf OscarTest.lg -k 10"

2. kCPD problem

• Extract Top-10 frequent patterns with diversidy in "Oscar" knowledge graph, with k = 10, core pattern is OscarTest, diversity threshold \lamda = 0.3: "./topkFreqDiv.sh -f Oscar.lg -cf OscarTest.lg -k 10 -t 0.3"

DATA

Expected graph format

v <node_id> <node_label>
v <node_id> <node_label>
...
e <node_id1> <node_id2> <edge_label>
e <node_id1> <node_id2> <edge_label>
...

In particular, it must first list all the graph nodes in ascending order of id, and then all the edges. Lines corresponding to nodes start with the letter v, while those corresponding to edges start with the letter e.

Synthetic graph generation

python2.7 graph_generator.py #ofNodes $degree $SizeofLabel $ProbabilityModle #GraphGenerationModel true
python2.7 graph_generator.py 1000 5 8 exp pa true

The scripts of labeled grpah gerneration (graph_generator.py) were from https://github.com/MSNTCS/graph-generator#readme, which called the Python interface from Snap.py. It is a Python interface for SNAP (Stanford Network Analysis Platform). SNAP is a general purpose, high performance system for analysis and manipulation of large networks. Additional information is available at http://snap.stanford.edu.