/NetVotes2

Primary LanguageGAPGNU General Public License v3.0GPL-3.0

NetVotes

  • Copyright 2017 Arinik Nejat

NetVotes is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation. For source availability and license information see licence.txt

Description

This repo is based on (and is complementary to) the following repo: https://github.com/CompNet/NetVotes This set of R scripts was designed to process signed graph algorithms, plot their graph results, and perform some statistical plots.

Data

Data representing the activity of the members of the European Parliament (MEPs) during the 7th term (from June 2009 to June 2014), as described in [MFLM'15]

Organization

Here are the folders composing the project:

  • Folder src: contains the source code (R scripts) to process signed graphs and plot their graph results.
  • Folder stats: contains the other source code (R scripts) to plot some statistics.
  • Folder out: contains the produced files (by 'src/run-netvotes.R').

Installation

  1. Install the R language
  2. Install the following R packages:
    • XML
      • in terminal, type: sudo apt-get install libxml2-dev
      • in R, type: install.packages("XML")
    • igraph: required (tested with version 1.0.1).
      • in terminal, type: sudo apt-get install libblas-dev liblapack-dev gfortran
      • type in R: install.packages("irlba")
      • type in R: install.packages("igraph")
    • iterators
    • foreach
    • doParallel
    • sourcetools
    • ggplot2 (if an error occurs like ‘ggplot2 is not available’, type in terminal ‘sudo apt-get install r-cran-ggplot2’)
    • treemap
  3. Install IBM Cplex 12.7.1
    • for ubuntu, type the following command:
      • sudo ./cplex_studio12.7.1.linux-x86-64.bin
        • the default installation location for education version is: /opt/ibm/ILOG/CPLEX_Studio1271
        • the default installation location for trial version is: /opt/ibm/ILOG/CPLEX_Studio_Community127/cplex/bin/x86-64_linux/
        • If you use trial version or want to change the default installation folder, update the corresponding variable CPLEX.BIN.PATH located in define-consts.R. Otherwise, no need to update it
  4. Install Open Java 1.8 (our ExCC jar file is compiled with that - TODO)
  5. Install OpenMPI for C++. Export LD_LIBRARY_PATH variable to the path where openmpi shared objects are located. For example, in my case: export LD_LIBRARY_PATH=/usr/lib/openmpi/lib:$LD_LIBRARY_PATH (if LD_LIBRARY_PATH is not set correctly, you will get ‘error when loading shared object ...’)
    • in terminal: sudo apt-get install gcc g++ openmpi-bin openmpi-doc libopenmpi-dev
    • check openmpi by typing in terminal: which mpicc
    • configure openmpi environment variables in your home by adding these 2 lines at the end of the ~/.bashrc file (change the 2 paths below depending on your installation):
      • export PATH=/usr/include/openmpi/:$PATH
      • export LD_LIBRARY_PATH=/usr/lib/openmpi/lib:$LD_LIBRARY_PATH
    • type in terminal: source ~/.bashrc
  6. (I am not sure if it is really needed) Install Boost C++ library
    • download the boost c++ library
    • decompress it on "tempdir"
    • run on the "tempdir": ./boostrap.sh
    • once the configuration succeds edit the file "project-config.jam" and add the "using mpi ;" at the end.
    • run in terminal: sudo ./b2 --target=shared,static --with=all -j2 install
    • run in terminal: sudo ldconfig
    • check if mpi .so files are installed correctly
      • run: ls /usr/local/lib
      • check if the files "libbost_mpi.a", "libbost_mpi.so" are there
  7. Install numpy and scipy for python
  8. Download this project from GitHub and unzip the archive.
  9. Go to the current project directory
  10. In terminal: chmod a+x exe/grapspcc and chmod a+x exe/kmbs
  11. configure src/run-netvotes.R file.
  12. configure LIBRARY.LOC.PATH in src/run-netvotes.R and stats/real-instances/main.R Initially, LIBRARY.LOC.PATH = .libPaths(). But, you can configure it (especially, if you downloaded all R libs in a specific dir). What is easier is to update your .libPaths() in /etc/R/Rprofile.site (i.e. I added my personal local lib dir by doing: .libPaths(c("~/R/R-library", .libPaths()))). Hence, I do not need to update the ".libPaths()" each R session
  13. configure/update CPLEX.BIN.PATH in src/define-consts.R
  14. configure MAX.G.SIZE located in stats/define-consts.R. I set to 150 because when graph-size increases (especially after 250), the execution of ExCC is getting slower.

Use

In order to replicate the experiments from the article, perform the following operations:

  1. Open the R console.
  2. Set the current projetct directory as the working directory, using setwd("my/path/to/the/project/SignedBenchmark").
  3. Run src/run-netvotes.R (you might first check the boolean variables)
  4. Run stats/main.R

Info

  • The variables we are using are global variables (src/run-netvotes.R, src/define-consts.R, ...). They will be used throughout the code.
  • Stats'lardaki plotlar icin: regarding "assessment for filtering": isolated node'lari kaldirmadan islem yapiyorum. Bununla birlikte, filtering versiyonda 2 algo sonculari karsilastirmasinda isolated node'lar membership bilgisinden cikariliyor.
  • in prepare.grasp.ils.parameter.set: g.size'a gore parameter set degisiyo => section 4 in [MRYL'17]
  • handling membership/nb.cluster variables for isolated nodes: we put all isolated nodes in the same cluster in the 'membership' variable. Their cluster no is CLU.NO.FOR.ALL.ISOLATED.NODES ==> in plots, they are visualized in white color.
  • stats’daki scriptin duzgun calismasi icin: input olarak verilen metodlarin hem filtresiz hem de filtreli versiyonlari ile birlikte olmasi lazim. Sadece filtreli versiyon varsa bug yaratir
  • ExCC’yi parallel calistirirken bug oluyo. Muhtemelen memory cok gerektiren bi algo eger graph-size yuksekse. O yuzden o algo’yu sequential modda calistir
  • plot’taki “node-id-enabled” option’u ile “imb-edge-contr” opsiyonu ayni anda kullanilmaz. Cunku ikisi de vertex.label kullaniyo. Vertex.frame kalinligi ile belki “imb-edge-contr” opsiyonu kullanilabilir ama su an igraph’ta o ozellik yok
  • plot’taki “imb-edge-contr” opsiyonu su an aktif degil. Cunku onu benim R’de kodlamam lazil. Yani vertex’lerin imbalance contribution’larini hesaplatmak gerek. Mario’nun yaptigini pek dogru degil gibi. Zaten eger R’de hesaplatirsam dige algo’lar icin de uygulamis olurum
  • Mario’nun algosu normalde parallel modda calistirabilriiz. Mpirun ile ayarlanabilir kolayca. Ama ben hic calistirmadim oyle.

To-do List

  • imbalance hesaplamasini genislet. Total.imbalance, neg.imbalance ve pos.imbalance
  • implement vertex imbalance contributions from membership file. It will be the same task that Mario has done. But, we will be able to use it for algorithms if we implement it in R. Also, I do not understand what Mario's code for this task. The result might be wrong in some cases.
  • Target types 2 tane. 1 tane olursa sikinti olur mu execution’da? Duzelt.
  • treemap’te isolated node’u (dikdortgen parca) beyaz ile goster

References

  • [MFLM'15] Mendonça, I.; Figueiredo, R.; Labatut, V. & Michelon, P. Relevance of Negative Links in Graph Partitioning: A Case Study Using Votes From the European Parliament, 2nd European Network Intelligence Conference (ENIC), 2015.
  • [MRYL'17] Mario Levorato, Rosa Figueiredo, Yuri Frota & Lúcia Drummond. Evaluating balancing on social networks through the efficient solution of correlation clustering problems, 2017