Machine Learning/Kaggle Social Network Contest/lit review

From Noisebridge
< Machine Learning | Kaggle Social Network Contest
Revision as of 21:30, 19 November 2010 by Jjhale (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

This page contains links to relevant articles and summaries of the papers.

[edit] Papers

[edit] Supervised Random Walks

  • title: "Supervised Random Walks: Predicting and Recommending Links in Social Networks"
  • authors: Lars Backstrom and Jure Leskovec
  • paper
  • Summary
    • develop an algorithm based on Supervised Random Walks
    • uses network structure info combined with node and edge level attributes to guide the walk
    • learn a function to weight edges s.t. random walker more likely to visit nodes to which new links will be created (equivalent to missing nodes for our application)
    • they develop a good training algorithm.
    • test it on a facebook network and on co-author network
    • compare to decision trees, logistic regression and unsupervised techniques.
Personal tools