This repository is the implementation of the paper MELOPPR: Software/Hardware Co-design for Memory-efficientLow-latency Personalized PageRank published at DAC'21.
- Python3 preferred
- Networkx package
MeLoPPR is tested on six graphs: citeseer, cora, pubmed, dblp, amazon, and youtube.
This repository contains the first three, and you may download the others from SNAP
- equation_validation.py: validates the stage decomposition and linear decomposition equations
- toy_example.py: validates the algorithm on the citeseer graph
- PPR.py: include the optimizations discussed in the paper (with memory and CPU time measuring code commented)
If you experience bugs, or have suggestions for improvements, please use the issue tracker to report them.
If this code has been useful to your research, please consider citing us:
BibTeX:
@inproceedings{meloppr,
title={MeLoPPR: Software/Hardware Co-design for Memory-efficientLow-latency Personalized PageRank},
author={Lixiang Li, Yao Chen, Zacharie Zirnheld, Pan Li, and Cong Hao},
booktitle={2021 58th ACM/IEEE Design Automation Conference (DAC'21)},
year={2021}
}