Tutorial of variant implementations of PageRank

User 124 | 4/24/2014, 6:12:56 AM

GraphLab feels great! We recently had a tutorial on programming GraphLab API for our students. Post here to share with the community and have more eyes reviewing the technical content. Besides a basic setup guide, we gave 5 different implementations for PageRank to show the flexibility of GraphLab:

  • Collect-driven flow. (original in the demoapps)
  • Scatter-driven flow.
  • Fixed iteration via vertex scheduling
  • Fixed iteration via engine scheduling
  • Directly simulating random walker

Those may not be used in production due to efficiency, but could be learning examples to explore the interfaces: what can/ cannot be done. If there are any errors or obsolete remarks under new version, please let me know.



User 6 | 4/24/2014, 7:38:47 AM

This is a great resource! We will add a link to your detailed tutorial from ours.


User 6 | 4/28/2014, 2:05:12 PM

p.s. A nice addition could be pgerank using the warp engine.