ecga.m

This matlab code on reed solomon and BCH encoding and different decoding algorithms. Also errors and

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% File: eCGA.m
%
% Description: The main loop of eCGA. 
%
% Author: Kumara Sastry
%
% Date: March 2007
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

clear;

% Get the user defined parameter values;
parameters = userDefinedParameters();

% Number of bits
ell = parameters.NumVariables;

% Population size
n = parameters.PopulationSize;

% Alphabet cardinalities
ranges(1:ell) = parameters.Cardinalities;

% Selection Method
selection = parameters.Selection;
selectionMethod = selection{1};
if(strcmpi(selectionMethod,'TRUNCATION'))
    truncParam = selection{2};
else
    tournSize = selection{2};
end

% Replacement stragegy: Default is generational replacement
replacement = parameters.Replacement;
replacementMethod = replacement{1};
if(strcmpi(replacementMethod, 'RTR'))
    windowSize = replacement{2};
end

% Convergence critieria parameters 
convergenceMethods = parameters.Convergence{1};
convergenceParameters = parameters.Convergence{2};

% See if any of the convergence criteria are set:
%   1. Variance of fitness is below a threshold (might not be useful if
%      the fitness is noisy)
%   2. If the population has converged to a single solution
%   3. If the best fitness is greater than the maximum fitness desired
%   4. If the generation no. is greater than the maximum number of
%      generations.
convergenceCriteria = strcmpi(convergenceMethods, {'MAXGEN', 'MAXFIT', 'FITVAR', 'POPVAR'});

% Maximum number of generations
if(convergenceCriteria(1))
    maxGen = convergenceParameters{1}; 
else
    maxGen = 10*ell;
end

% Maximum fitness value desired (e.g., fitness of the optimal solution.
if(convergenceCriteria(2))
    maxFit = convergenceParameters{2}; 
else
    maxFit = inf;
end

% Threshold for fitness variance below which the population is assumed to
% be converged
if(convergenceCriteria(3))
    maxFitVar = convergenceParameters{3}; 
else
    maxFitVar = 0.0;
end

% Threshold for the unique number of individuals in the converged
% population.
if(convergenceCriteria(4))
    maxPopVar = convergenceParameters{4}; %
else
    maxPopVar = 1;
end


%--------------------------------------------------------------------------
% Start with eCGA
%--------------------------------------------------------------------------

% Initialize the population randomly
population = (rand(n,ell) > 0.5);

% Evaluate the fitness of the initial population
fitness = evaluateIndividuals(ranges, population);

% Set the convergence flag to false
isConverged = false;
gen = 0;
[bestFit, bestID] = max(fitness);
bestSolution = population(bestID,:);

% While the convergence criteria is not met
while ~isConverged,
    
    % Perform tournament selection and determine the mating pool
    if(strcmpi(selectionMethod,'TRUNCATION'))
        winners = truncationSelection(truncParam, fitness);
    else
        winners = tournamentSelection(tournSize, fitness);
    end
    
    % Build the marginal product model
    mpm = buildModel(ranges, population(winners,:));

    % Sample the new population using the MPM
    newPopulation = sampleModel(mpm, n);

    % Evaluate the fitness of the new individuals
    newFitness = evaluateIndividuals(ranges, newPopulation);

    % Determine the current best solution and fitness
    [currentBestFit, bestID] = max(fitness);
    currentBestSolution = population(bestID, :);

    % If the current best is better than the best-so-far replace the
    % best-so-far solution with the current best
    if(currentBestFit > bestFit)
        bestFit = currentBestFit;
        bestSolution = currentBestSolution;
    end

    % Replacement Strategies:
    %   1. Restricted tournament replacement
    %   2. Elitist Lambda+Mu Replacement
    %   3. Generational replacement
    if(strcmpi(replacementMethod,'RTR'))
        [population, fitness] = restrictedTournamentReplacement(population, fitness, newPopulation, newFitness, windowSize);
    elseif (strcmpi(replacementMethod,'ELITIST'))
        [population, fitness] = elitistReplacement(population, fitness, newPopulation, newFitness);
    else
        % DEFAULT: Generational replacement strategy
        population = newPopulation;
        fitness = newFitness;
    end
    
    % See if any of the convergence criteria are met:
    if((gen > maxGen)  || (bestFit >= maxFit) || (var(fitness) < maxFitVar) || (size(unique(population,'rows'),1) <= maxPopVar))
        isConverged = true;
    end
    
    gen = gen+1;
    
    % Display the model, best-so-far solution and its fitness
    displayResults(gen, mpm, bestSolution, bestFit);
end