fitscalingprop.m

遗传算法工具包

function expectation = fitscalingprop(scores,nParents)
%FITSCALINGPROP Proportional fitness scaling.
%   EXPECTATION = FITSCALINGPROP(SCORES,NPARENTS) calculates the EXPECTATION
%   using the SCORES and number of parents NPARENTS. The least sophisticated 
%   fitness scaling strategy, expectation is simply proportional to the raw 
%   fitness score. This strategy has weaknesses when raw scores are not in a
%   "good range".
%
%   Example:
%   Create an options structure using FITSCALINGPROP as the fitness scaling
%   function
%     options = gaoptimset('FitnessScalingFcn',@fitscalingprop);
%

%   Copyright 2004 The MathWorks, Inc. 
%   $Revision: 1.7 $  $Date: 2004/01/16 16:51:40 $

% Rotate scores around their mean because we are minimizing so lower scores
% must yield a higher expectation.
scores = 2 * mean(scores) - scores;

% Negative expectations are simply not allowed so here we make sure that
% doesn't happen by sliding the whole vector up until everything is
% non-negative.
m = min(scores);
if(m < 0)
    scores = scores - m;
end

% Normalize: expectation should sum to one.
expectation = nParents * scores ./ sum(scores);