robink87 / edgeppr

Edge-Weighted Personalized PageRank

This release contains implementations from paper "Edge-Weighted Personalized PageRank: Breaking A Decade-Old Performance Barrier" (KDD 2015).

Here we use ObjectRank on the DBLP dataset as an illustration.

Requirements

• numpy (>= 1.6.2)
• scipy (>= 0.10.1)

Usage

Download the preprocessed data from here, put it under this folder and decompress it:

tar xvf data.tar.gz

Run the following commands to execute the experiments for query answering and learning to rank with model reduction method:

script/answerquery.sh
script/learnrank.sh

After the execution, the experiemental results will be generated in the result folder.

Comming soon

• Code for preprocessing