RT_replacement

 [NewPop,NewFunVal] = RT_replacement(Pop,SelPop,SampledPop,FunVal,SelFunVal,SampledFunVal,replacement_params)
 RT_replacement            Creates a new population (NewPop) applying restricted tournament replacement. For each
                           newly generated individual,  first a subset  of (window) individuals from
                           the current population is selected. The individual genotypically closest (square
                           difference) to the candidate individual is found. The individual with highest evaluation
                           remains in the population. 
                           This method is appropriate for single objective optimization.
 INPUTS 
 Pop:                                 Current population
 SelPop:                              Current selected population
 SampledPop:                          Population sampled from the probabilistic model
 CurrentFunVal:                       A matrix of function evaluations, one vector of m objectives for each individual
 OUTPUTS
 NewPop                        : New Population
 NewFunVal                     : Evaluations of the new population
 window = replacement_params{1}(1): Subset of solutions considered in the comparison. 

 Last version 8/26/2008. Roberto Santana (roberto.santana@ehu.es)

0001 function [NewPop,NewFunVal] = RT_replacement(Pop,SelPop,SampledPop,FunVal,SelFunVal,SampledFunVal,replacement_params)
0002 % [NewPop,NewFunVal] = RT_replacement(Pop,SelPop,SampledPop,FunVal,SelFunVal,SampledFunVal,replacement_params)
0003 % RT_replacement            Creates a new population (NewPop) applying restricted tournament replacement. For each
0004 %                           newly generated individual,  first a subset  of (window) individuals from
0005 %                           the current population is selected. The individual genotypically closest (square
0006 %                           difference) to the candidate individual is found. The individual with highest evaluation
0007 %                           remains in the population.
0008 %                           This method is appropriate for single objective optimization.
0009 % INPUTS
0010 % Pop:                                 Current population
0011 % SelPop:                              Current selected population
0012 % SampledPop:                          Population sampled from the probabilistic model
0013 % CurrentFunVal:                       A matrix of function evaluations, one vector of m objectives for each individual
0014 % OUTPUTS
0015 % NewPop                        : New Population
0016 % NewFunVal                     : Evaluations of the new population
0017 % window = replacement_params{1}(1): Subset of solutions considered in the comparison.
0018 %
0019 % Last version 8/26/2008. Roberto Santana (roberto.santana@ehu.es)
0020 
0021 
0022 window = cell2num(replacement_params{1}(1)); 
0023 PopSize = size(Pop,1);
0024 SelPopSize = size(SelPop,1);
0025 SampledPopSize = size(SampledPop,1);
0026 
0027 NewPop = Pop;
0028 NewFunVal = FunVal;
0029 
0030 tonorm = repmat((max(Pop) - min(Pop)),window,1);
0031 for i=1:PopSize,
0032   auxperm = randperm(PopSize);
0033   candidates =  auxperm(1:window);
0034   auxsubpop = repmat(SampledPop(i,:),window,1);
0035    
0036 
0037   aux = ((auxsubpop - NewPop(candidates,:))./ tonorm);
0038   
0039   [minval,minpos] = min(sum(aux'.^2));
0040   closest_sol = candidates(minpos);
0041    
0042    if SampledFunVal(i) >= NewFunVal(closest_sol)
0043      NewPop(closest_sol,:) = SampledPop(i,:);
0044      NewFunVal(closest_sol) = SampledFunVal(i);     
0045    end
0046 end
0047 
0048 return
0049  
0050  
0051