Skip to content

fuvty/LessMine

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LessMine

This is a python implementation of LessMine: Reducing Sample Space and Data Access for Dense Pattern Mining

Fu T, Wan Z, Dai G, et al. LessMine: Reducing Sample Space and Data Access for Dense Pattern Mining[C]//2020 IEEE High Performance Extreme Computing Conference (HPEC). IEEE, 2020: 1-7.

NOTE: currently, the released python implementation is not the complete version because:

  1. It only support clique mining now. However, the method can be applied to any dense pattern as defined in the paper. For further details, please refer to our paper ( e.g. Section VI. )
  2. The optional locality-aware partition is not applied

Problem setup

Given a query pattern Q and a target graph G, estimate how many Qs are there in G. For LessMine, the query pattern Q should be a dense pattern (the pattern that has at least one node connected to all other nodes).

Note that the number of Qs is defined as the edge-induced subgraph of G, a.k.a. monomorphism of G.

Core Idea

The essense of LessMine is ConcurrentSample and UniformClose . The function LessMine_core in LessMine.py provides a complete vision of how LessMine works. One can refer to the paper to understand how LessMine reduces the sample space and data access for dense pattern mining.

Usage

The file LessMine.py provides the main user interface called LessMine_main() and the main() function is an example of how to use it

LessMine_main takes three args: target, query and num_estimator

  • target_filename is a string. It indicate the filename of the edge list of the target graph

  • query is nx.Graph, the networkx instance of query. Note that for the nx.Graph instance with n nodes, the node set should be ${0,1,\dots,n-1}$, and the first node of any edge is always smaller than the second

  • num_estimator is the number of estimators used to estimate the number of Q, one can adjust the number according to the time-error budget. Typically, larger G and Q requires more estimators.

    We use 1E4 estimators on ego-Facebook(#E = 88234, #V = 4039) for 4-clique counting to get a approximately 10% error-bound

//  Created by Tianyu Fu on 2021/5/24.
//  Contributed by members of NICS Lab of Tsinghua University

//  Contact: fty22@mails.tsinghua.edu.cn

About

The code implementation of HPEC20 paper *LessMine: Reducing Sample Space and Data Access for Dense Pattern Mining*

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages