pso.rar_Griewank_matlab griewank_pso ackley_pso rosenbrock_粒子群 G

% particle swarm optimization algorithm % function [fxmin, xmin] =pso % clc; clear; % Parameter for swarm % tic; SwarmSize= 50; % the number of particles or individuals in Swarm Dim = 100 ; % the dimension of solution SwarmX = zeros(SwarmSize, Dim); % particle's position matrix, which is a SwarmSize by Dim matrix SwarmV = zeros(SwarmSize, Dim); % particle's velocity matrix, which is a SwarmSize by Dim matrix range=ones(2,Dim); IterMax = 2000; % maximum iteration number fname = 'griewank'; % The funciton to be optimized runtime=30; Globalmins=zeros(1,runtime); pos=zeros(15,runtime); for r=1:runtime XRange(1,:)=range(1,:)*(-600); XRange(2,:)=range(2,:)*600; % matrix of 2 by Dim , which specifize the range of each variables Vmax =4; % velocity range,i.e. [v_min, v_max] Pibest = zeros(SwarmSize, Dim); % the individual best position Pgbest = zeros(1, Dim); % the global best position fhistory = [ ]; % variable to record the best fitness during iteration % Parameter for algorithm % c1 = 2; c2 = 2; % parameter for c1 and c2 % innertia weight using linear decreasing w_max = 0.9; w_min = 0.4; bflag = 1; % method flags to process particle's position that exceed XRange: 0--no limit, 1: limit vflag = 1; % method flags to process particle's velocity that exceed Vmax: 0--no limit, 1: limit fmin = 0; % variable to strore the gobal minimum fitness % ==========Initialization============== % initialize particle's position within VarRange SwarmX = rand(SwarmSize, Dim); for i = 1:Dim SwarmX(:,i) = XRange(1,i)+ (XRange(2, i) - XRange(1,i))*SwarmX(:, i); end % intialize particle's velocity SwarmV = -Vmax+ 2*Vmax*rand(SwarmSize,Dim); % evalute the fitness of particles fSwarm = feval(fname, SwarmX); % initialize Pibest and Pgbest [fg, idx] = min(fSwarm); Pgbest = SwarmX(idx, :); Pibest =SwarmX; fPibest = fSwarm; % initially, each particle's best position are set equal to fSwarm fmin = fg; fhistory = [fhistory; fmin]; % recorde the best fitness, here the best position is not recorde, if necessary, one can using a variable to recorde it t = 0; % counter for iteration % =========== end initialization ========== SwarmMaskmin = repmat( XRange(1,:), SwarmSize,1); SwarmMaskmax = repmat( XRange(2, :), SwarmSize,1); % ====== start pso loop ======= while t < IterMax % update iteration counter t = t+1; w = w_max - t*(w_max - w_min)/IterMax; % calculate new inertia weight % update particle's velocity A = repmat(Pgbest, SwarmSize, 1); SwarmV = w*SwarmV + c1*rand(SwarmSize, Dim).*(Pibest - SwarmX)+c2*rand(SwarmSize,Dim).*(A - SwarmX); % check SwarmV wheter is in [-Vmax, Vmax] if vflag ==1 chRows = SwarmV > Vmax; SwarmV(find(chRows)) = Vmax; chRows = SwarmV < -Vmax; SwarmV(find(chRows)) = -Vmax; end % update particle's position SwarmX = SwarmX + SwarmV; % Comfirm particle's position, such that it is in VarRange % method 0 : no position limiting % method 1: saturation at limiting % method 2: bounce off limit % we adopt method 1 here if bflag ==1 minSwarm_throw = SwarmX <= SwarmMaskmin; minSwarm_keep = SwarmX > SwarmMaskmin; maxSwarm_throw = SwarmX >= SwarmMaskmax; maxSwarm_keep = SwarmX < SwarmMaskmax; SwarmX = ( minSwarm_throw.*SwarmMaskmin ) + ( minSwarm_keep.*SwarmX ); SwarmX = ( maxSwarm_throw.*SwarmMaskmax ) + ( maxSwarm_keep.*SwarmX ); end % evaluate new fitmess of particles fSwarm = feval(fname, SwarmX); % update Pibest and Pgbest chRows = fSwarm < fPibest; fPibest(find(chRows)) = fSwarm(find(chRows)); Pibest(find(chRows), :) = SwarmX(find(chRows), :) ; [fg, idx] = min(fPibest); % here has a question: [fg, idx] = min(fSwarmX) % update gbest history if fg < fmin Pgbest = Pibest(idx,:); % or SwarmX fPgbest = fg; end fmin = min(fPibest); % fhistory = [fhistory; fmin]; fprintf('t=%d fmin=%g\n',t,fmin); if (t==200) pos(1,r)=fmin; end; if(t==400) pos(2,r)=fmin; end; if(t==600) pos(3,r)=fmin; end; if(t==800) pos(4,r)=fmin; end; if(t==1000) pos(5,r)=fmin; end; if(t==1200) pos(6,r)=fmin; end; if(t==1400) pos(7,r)=fmin; end; if(t==1600) pos(8,r)=fmin; end; if(t==1800) pos(9,r)=fmin; end; if(t==2000) pos(10,r)=fmin; end; if(t==2200) pos(11,r)=fmin; end; if(t==2400) pos(12,r)=fmin; end; if(t==2600) pos(13,r)=fmin; end; if(t==2800) pos(14,r)=fmin; end; if(t==3000) pos(15,r)=fmin; end; end % ====== end pso loop ====== % output the final result end; fprintf('%d\n',mean(pos,2)); save all