cranky-cyborg / game-theory-voting-system Goto Github PK

README 
Ronald L. Rivest and Emily Shen
March 29, 2010

This is about material posted in the directory:
   http://people.csail.mit.edu/rivest/gt

This is the README file for the code and experimental results
for the GT Voting System, as described in:
   "An Optimal Single-Winner Preferential Voting System 
    Based on Game Theory"
   by Ronald L. Rivest and Emily Shen
This paper is posted on Rivest's publication page:
   http://people.csail.mit.edu/rivest/publications.html
   as well as being given in this directory. (The 
   former may be more up-to-date.)

This directory contains the following files:

README	    This file.

vs.py	    The main voting system program, including
            code for reading election profiles, computing
            winners according to various voting systems,
            and comparing various voting systems to each other.
            Written in python (tested in python 2.6).
	    It requires game_cvxopt.py .

	    It can be run on an election profile, as
	        python vs.py data/ex_15.txt
	    or used to compare various voting systems
	    on simulated data, as
                python vs.py -compare

game_cvxopt.py

            This is a python module that solves two-person
            zero-sum games, using the CVXOPT package
 	    (which must be loaded separately; see
                http://abel.ee.ucla.edu/cvxopt/
	    ). It contains an "LP"-based solver,
            which returns *some* optimal mixed
            strategy, and also a "QP"-based solver,
            which returns the unique *balanced* optimal
            mixed strategy (minimizing sum of squares).
	    This module is called by vs.py.

data	    This is a subdirectory containing various
            sample election profiles as ".txt" files,
            and also the corresponding margin matrices
            (as ".margin" files)

experiment5 This directory contains the code and output
            for the experiment reported in our paper.

gtlp.m	    Matlab code for solving two-person zero-sum
            game using LP solver.  (Can be run on data
            available in a ".margin" file.)  Returns
            *some* optimal mixed strategy.  Not used
	    for the results in our paper, but might
	    be of interest to some folks.

gtqp.m      Matlab code for solving two-person zero-sum
            game using QP solver. (Can be run on data
            available in a ".margin" file.)  Returns
            the unique optimal *balanced* mixed
	    strategy (minimizing sum of squares).
	    Not used for the results in our paper,
	    but might be of interest to some folks.

gt.pdf	    The paper describing the GT voting system.
            (May not be most recent.)