DualIso : Scalable Subgraph Pattern Matching On Large Labeled Graphs SUPPLEMENT

This supplement to the paper provides all figures and substantial code listings which would not fit into the paper due to page limitations. Roadmap: The supplement is organized in the same manner as the main paper i.e. the section names in the main paper corresponds to the same section in the supplement for the easy reference. Note: In many cases, text from the main paper is included and or extended in order to provide context for the figures, tables, and code listings contained in this supplement. The figure numbers may get changed as some more figures are added in this supplement for detailed explanation. A. Graph Pattern Matching The goal of graph pattern matching is to find all subgraphs of a large graph, called the data graph, that are similar to a query graph structurally and semantically. To consider why this might be useful, look again at the graph displayed in Figure 1. Suppose a user of Facebook has just moved to Lyon, France from the United States and is looking for someone to accompany him to the U2 concert there. Thus, he wants to find out if any of his friends are friends with a person who lives in Lyon and who likes U2. Given that Facebook can be modeled as a graph like Figure 1, this question can be modeled as a graph query, as shown in Figure 2. In this query, the capital letters represent variables that the user would like filled in. In this case, using the graph in Figure 1, the query would return X = John and Y = Jim. In fact, Facebook Graph Search2 attempts to allow users to do just that, though at the moment it only has limited querying capabilities.