Archive for the ‘Graph Data Management’ Category

Sketch-based Approach for Graphs

We discussed these two papers about sketch-based approach today:

This key of sketch-based approach is finding the landmarks by random sampling for every node in a graph. Some comments:

  • The assumption for sketch-based approach is the nodes are uniformly selected. This assumption is only suitable for random graphs. For scale-free graphs, nodes have distinct priority, so the uniform sampling technique may be not the case;
  • Making whatever incremental computation based on sketch-based approach seems an explored research topic;
  • The combination of sketch-based approach and random walk may generate some very fast algorithms for PageRank/SimRank on large graphs.
Categories: Graph Data Management Tags: