Code Contributors: Xiaodong He

If you have any questions, feel free to contact me. My emails are: [email protected]

Please cite our paper if you choose to use our code.

@inproceedings{wei2018topppr,
  title        = {Topppr: top-k personalized pagerank queries with precision guarantees on large graphs},
  author       = {Wei, Zhewei and
                  He, Xiaodong and
                  Xiao, Xiaokui and
                  Wang, Sibo and
                  Shang, Shuo and
                  Wen, Ji-Rong},
  booktitle    = {Proceedings of the 2018 International Conference on Management of Data},
  pages        = {441--456},
  year         = {2018},
  organization = {ACM}
}

Single Source Personalized PageRank

Tested Environment

• Ubuntu
• C++ 11
• GCC 4.8

Preparation

Place the prepared data [file].txt in the ./dataset folder and create the subfolder [file] with the same name in the ppr-answer folder. Note that the first row of data is the node size, and each row is the information of each edge [outNode] [inNode]. test.txt is example.

Compile

sh compile.sh

Parameters

./TopPPR --algo <algorithm> [options]

• algo:
  • GEN_QUERY: generate query node file
  • GEN_GROUND_TRUTH: generate query node groundtruth by multi-thread power method
  • TopPPR: TopPPR query
• options
  • -d <dataset>
  • -k <top k>
  • -n <query size>
  • -r <precision threshold>
  • -err <max error (defalt 10^-10)>
  • -a <alpha (default 0.2)>

Generate queries

Generate query files for the graph data.

• Example:

$ sh gen_query.sh

Generate groundtruth of query nodes

• Example:

$ sh gen_groundtruth.sh

Exact Query

• Example:

./TopPPR -d test -algo TopPPR -r 1 -n 50 -k 5

Approximate Query

• Example:

./TopPPR -d test -algo TopPPR -r 0.4 -n 50 -k 5