Problem with Output function in Genetic Algorithm (GA)

Question

Marco Milanesi 2021-3-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/779192-problem-with-output-function-in-genetic-algorithm-ga

关闭： Marco Milanesi 2023-8-4

This is my main script, I have a problem in the output function which have to save a mat file with the hystory of each generation. The "Output function" (@myfun in the code) that I have written save the hystory of the first and the second ga but in the third ga it only saves the first and the last value.

Main function

clear all, close all, clc
dt = 0.001;
popsize = 10;
MaxGenerations = 10;
s = tf('s');
G = (1/(1+5*s))*exp(-s)
rng(1,'twister') % for reproducibility
population_PID = rand(popsize,3);
population_I_PD = rand(popsize,3);
population_PI_D = rand(popsize,3);
%% PID
options_PID = optimoptions(@ga,'PopulationSize',popsize,'MaxGenerations',MaxGenerations,'InitialPopulation',population_PID,'OutputFcn',@myfun_pid);
[x_PID,fval_PID,exitflag,output,population_PID,scores] = ga(@(K)pidtest(G,dt,K),3,-eye(3),zeros(3,1),[],[],[],[],[],options_PID);
fval_PID
%% I-PD
options_I_PD = optimoptions(@ga,'PopulationSize',popsize,'MaxGenerations',MaxGenerations,'InitialPopulation',population_I_PD,'OutputFcn',@myfun_ipd);
[x_I_PD,fval_I_PD,exitflag,output,population_I_PD,scores] = ga(@(K)ipdtest_gapid(G,dt,K),3,-eye(3),zeros(3,1),[],[],[],[],[],options_I_PD);
fval_I_PD
%% PI-D
options_PI_D = optimoptions(@ga,'PopulationSize',popsize,'MaxGenerations',MaxGenerations,'InitialPopulation',population_PI_D,'OutputFcn',@myfun_pi_d);
[x_PI_D,fval_PI_D,exitflag,output,population_PI_D,scores] = ga(@(K)dpitest(G,dt,K),3,-eye(3),zeros(3,1),[],[],[],[],[],options_PI_D);
fval_PI_D

My fun PID

All the output function are the same but only in the last ga that I run it doesn't save in the correct way.

function [state, options,optchanged] = myfun_pid(options,state,flag)
persistent history_pid 
persistent cost_pid
optchanged = false;
switch flag
 case 'init'
        history_pid(:,:,1) = state.Population;
        cost_pid(:,1) = state.Score;
    case {'iter','interrupt'}
        ss = size(history_pid,3);
        history_pid(:,:,ss+1) = state.Population;
        cost_pid(:,ss+1) = state.Score;
    case 'done'
        ss = size(history_pid,3);
        history_pid(:,:,ss+1) = state.Population;
        cost_pid(:,ss+1) = state.Score;
        save history_pid.mat history_pid cost_pid
end

hystory_pid.mat

My fun PI-D

function [state_PI_D, options_PI_D,optchanged] = myfun_pi_d(options_PI_D,state_PI_D,flag_PI_D)
persistent history_dpi 
persistent cost_dpi
optchanged = false;
switch flag_PI_D;
 case 'init'
        history_dpi(:,:,1) = state_PI_D.Population;
        cost_dpi(:,1) = state_PI_D.Score;
    case {'iter','interrupt'}
        ss = size(history_dpi,1);
        history_dpi(:,:,ss+1) = state_PI_D.Population;
        cost_dpi(:,ss+1) = state_PI_D.Score;
    case 'done'
        ss = size(history_dpi,1);
        history_dpi(:,:,ss+1) = state_PI_D.Population;
        cost_dpi(:,ss+1) = state_PI_D.Score;
        save history_pi_d.mat history_pi_d cost_pi_d
end