实现了极值动力学寻优的特点_有效的解决了关于函数寻优_matlab

%----------------------------------------- % Approaching-based Extremal Optimization % 没有最优解引导的随机单向变异 % name: M_AEO.m % by zhuming Dec 2, 2008 %----------------------------------------- function [fit_aver,SR,time_aver,best_fit,best_aver,generation_aver,fitness_best] = EO(er,Label_function); clear fit_aver SR time_aver best_fit best_aver generation_aver fitness_best % 不同的函数设置不同的参数区间 if Label_function == 2 min_v = -5.12; max_v = 5.12; range_mut = 0.05; end pop_size = 60; % 种群大小 indi_length = 2; % 个体长度，组元（基因）数 len_v_r = max_v - min_v; % 参数区间长度 generation_max = 1000; % 最大进化代数 fail_time = 0; run_max = 10; % 运行次数 run_gen = zeros(1,run_max); % 每次运行的收敛代数 run_time = zeros(1,run_max); % 每次运行的时间 fit_run = zeros(run_max,generation_max); % 每次运行的时间 best_run = zeros(1,run_max); bool_break = 0; for r = 1:run_max time = clock; S = len_v_r*(rand(pop_size,indi_length)-0.5); % 初始化种群 S_best = zeros(1,indi_length); % 当前最优解 fit = zeros(1,pop_size); % 适应度函数值 fit_best = 0; % 最优适应度值 fitness = zeros(1,generation_max); obj = zeros(1,pop_size); for i = 1:pop_size % 计算个体的适应度函数值 x1 = S(i,1); x2 = S(i,2); obj(i) = fitness_value(x1,x2,Label_function); fit(i) = obj(i); end indx_best = 0; fit_best = 0; for i = 1:pop_size if fit_best < fit(i) indx_best = i; fit_best = fit(i); end end S_best = S(indx_best,:); % 随机选择一个个体作为最优个体，此处选择第一个体 fitness(1) = fit_best; best_run(r) = fit_best; for g = 1:generation_max % 进化过程 S_temp = zeros(indi_length+2,indi_length); % 该个体可能的下一代个体 obj_temp = zeros(1,indi_length+2); fit_temp = zeros(1,indi_length+2); for i = 1:pop_size % 对每个个体进化操作 for j = 1:indi_length+2 S_temp(j,:) = S(i,:); end for j = 1:indi_length % 对个体的每个基因单独变异，并计算其适应度函数值 % Uniform Mutation if rand > 0.5 temp = S_temp(j,j) + range_mut*rand*abs(max_v - S_temp(j,j)); while temp < min_v || temp > max_v temp = S_temp(j,j) + range_mut*rand*abs(max_v - S_temp(j,j)); end S_temp(j,j) = temp; else temp = S_temp(j,j) - range_mut*rand*abs(S_temp(j,j) - min_v); while temp < min_v || temp > max_v temp = S_temp(j,j) - range_mut*rand*abs(S_temp(j,j) - min_v); end S_temp(j,j) = temp; end end for j = 1:indi_length+2 x1 = S_temp(j,1); x2 = S_temp(j,2); obj_temp(j) = fitness_value(x1,x2,Label_function); fit_temp(j) = obj_temp(j); end % 对可能解的适应度值进行排序，找出最差组元并将其对应的解代替当前解 for j = 1:indi_length+2 % 取多个最差中的最优者 if fit(i) < fit_temp(j) % 将最差组元Xiw对应的个体Siw替换Si fit(i) = fit_temp(j); S(i,:) = S_temp(j,:); end end if fit(i) > fit_best % 更新最优解 S_best = S(i,:); fit_best = fit(i); end fitness(g) = fit_best; best_run(r) = fit_best; end if (1-fit_best) < er run_gen(r) = g; bool_break = 1; break; end end if bool_break == 1 bool_break = 0; else run_gen(r) = g; fail_time = fail_time + 1; end run_time(r) = etime(clock,time); fit_run(r,:) = fitness(:); end % 结果输出 best_fit = max(best_run); % 最佳值 temp_A = fit_run; % 适应度为0者设为最大 for i = 1:run_max for j = 1:generation_max if temp_A(i,j) == 0 temp_A(i,j) = best_fit; end end end fit_aver = mean(temp_A,1); % 平均适应值 best_aver = mean(best_run); % 平均最优适应度值 time_aver = mean(run_time); % 平均运行时间 SR = run_max - fail_time; % 成功率 generation_aver = mean(run_gen); % 平均代数 best_indx = generation_max; for i = 1:run_max for j = 1:generation_max if fit_run(i,j) == best_fit & best_indx > i best_indx = i; break; end end end fitness_best = fit_run(best_indx,:); % 最优值的适应度过程 for i = 1:generation_max if fitness_best(i) == 0 fitness_best(i) = best_fit; end end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% if Label_function == 2 for i = 1:run_max best_run(i) = 1/best_run(i) - 1; end for i = 1:run_max for j = 1:generation_max temp_A(i,j) = 1/temp_A(i,j) - 1; end end end best_aver = mean(best_run); % 平均最优适应度值 best_fit = min(best_run); % 最佳值 fit_aver = mean(temp_A,1); % 平均适应值 fitness_best = temp_A(best_indx,:); % 最优值的适应度过程