Harnessing Graph Data: Problems, Techniques and Applications

Presented by Yuan Fang (University of Illinois) – Harnessing Graph Data: Problems, Techniques and Applications

Problem: classification
Challenge: limited context.

Method: Dirichlet model for graph based classification
Model both similarity and confidence

Random walk interpretation
Result: both graph -> lexical graph -> click graph -> flat data

Random walk on patio-temporal graph

---

Problem: Semantic-awre ranking on heterogeneous graph
Method: Metagraph representations
Challenges:

• One semantic ranking can have many metagraphs
• Metagraphs can have varying weights
• Not mutually exclusive

Too long, not record...