Parallel personalized pagerank on dynamic graphs
Personalized PageRank (PPR) is a well-known proximitymeasure in graphs. To meet the need for dynamic PPRmaintenance, recent works have proposed a local updatescheme to support incremental computation. Nevertheless,sequential execution of the scheme is still too slow for highspeedstream processing. T...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Institutional Knowledge at Singapore Management University
2017
|
| Subjects: | |
| Online Access: | https://ink.library.smu.edu.sg/sis_research/3900 https://ink.library.smu.edu.sg/context/sis_research/article/4902/viewcontent/p93_guo.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Singapore Management University |
| Language: | English |
| Summary: | Personalized PageRank (PPR) is a well-known proximitymeasure in graphs. To meet the need for dynamic PPRmaintenance, recent works have proposed a local updatescheme to support incremental computation. Nevertheless,sequential execution of the scheme is still too slow for highspeedstream processing. Therefore, we are motivated todesign a parallel approach for dynamic PPR computation.First, as updates always come in batches, we devise a batchprocessing method to reduce synchronization cost among everysingle update and enable more parallelism for iterativeparallel execution. Our theoretical analysis shows that theparallel approach has the same asymptotic complexity asthe sequential approach. Second, we devise novel optimizationtechniques to e↵ectively reduce runtime overheads forparallel processes. Experimental evaluation shows that ourparallel algorithm can achieve orders of magnitude speedupson GPUs and multi-core CPUs compared with the state-ofthe-artsequential algorithm. |
|---|