mining algorithms that allows for label and structural mismatches in the isomorphisms are useful in many real world scenarios Problem Statement Given a graph database label match cost matrix label mismatch threshold ID: 802832
Download The PPT/PDF document "Motivation Frequent subgraph" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Motivation
Frequent
subgraph
mining algorithms that allows for label and structural mismatches in the
isomorphisms
are useful in many real world scenarios.
Problem Statement
Given a graph database, label match cost matrix, label mismatch threshold
, structural mismatch threshold and minimum support . Find frequent approximate patterns.
Challenges
Subgraph
isomorphism is NP-Complete. Label and structure relaxations are even harder.
Number of
isomorphisms are exponential.
Thesis Contributions
CMDB Miner : Mining representative patterns from a single large graph.LMiner : Mines approximate patterns with bounded label cost and ZERO structural cost.AMiner: Mines approximate patterns with bounded label and structural costs.
Search Space Pruning
Label for each pattern vertex to prune its potential mappings in the database.
K-hop Label, Neighbor Concatenated Label
Representative Support of Pattern
Handling Structural Mismatches
Experiments
References
Pranay
Anchuri
,
et.al
Approximate Graph Mining with Label
Costs
.
, SIGKDD 2013
Pranay
Anchuri, et. al Infrastructure Pattern Discovery in Configuration Management Databases via Large Sparse Graph Mining., ICDM 2011
Delete edges from the pattern and use
LMiner to find matches.Challenges : Choosing the edges to delete and the order of deletion.
Algorithms for Mining Frequent Approximate Patterns from Graph DatabasesPranay Anchuri1,2, Mohammed Zaki1,2, Omer Barkol3, Shahar Golan31Department of CS, Rensselaer Polytechnic Institute, Troy, New York2Qatar Computing Research Institute, Doha, Qatar3HP Labs, Technion City, Israel.
A
B
A
C
A
B
C
C
Label Cost = Cost(A, C)
Structural Cost = 1