nsga2.m

%% clear all workspace variables
clear ;

%% this is where all the algorithm parameters are
addpath(genpath('./input_data'));   
% this is where all the problems are defined
addpath(genpath('./problemdef'));   
% this is where all the legacy rng related stuffs are.
% THIS IS NOT VECTORIZED, SO DO NOT USE, SLOW !!!
addpath(genpath('./rand'));         

%% global variables that may be used here
global popsize ;
global nreal ;
global nbin ;
global nbits ;
global nobj ;
global ncon ;
global ngen ;

%% load algorithm parameters
load_input_data('input_data/zdt4.in');
pprint('\nInput data successfully entered, now performing initialization\n\n');

%% for debugging puproses 
% global min_realvar ;
% global max_realvar ;
% popsize = 24 ;
% nreal = 3 ;
% ngen = 400 ;
% min_realvar = min_realvar(1:nreal);
% max_realvar = max_realvar(1:nreal);
% global min_binvar ;
% global max_binvar ;
% nbin = 2;
% nbits = [3;3];
% min_binvar = [0;0];
% max_binvar = [5;5];

if(nreal > 0)
    obj_col = nreal + 1 : nreal + nobj ;
elseif(nbin > 0)
    obj_col = sum(nbits) + 1 : sum(nbits) + nobj ;
end

%% this is the objective function that we are going to optimize
obj_func = @zdt4 ;

%% allocate memory for pops
if(nreal > 0)
    child_pop = zeros(popsize, nreal + nobj + ncon + 3);
    mixed_pop = zeros(2 * popsize, nreal + nobj + ncon + 3);
elseif(nbin > 0)
    child_pop = zeros(popsize, sum(nbits) + nobj + ncon + 3);
    mixed_pop = zeros(2 * popsize, sum(nbits) + nobj + ncon + 3);
end

%% you need to warm-up the cache if you need to do profiling
% for k = 1:50000
%     tic(); elapsed = toc();
% end

%% start nsga2
tic;
% initialize population
parent_pop = initialize_pop(90);
pprint('Initialization done, now performing first generation\n\n');
parent_pop = evaluate_pop(parent_pop, obj_func);
parent_pop = assign_rank_and_crowding_distance(parent_pop);

do_plot = false ;
% plot the pareto front
if(do_plot); show_plot(1, parent_pop, false, [1 2 3]); end;

do_save = false ;
% save the current pop
if(do_save); save_pop(1, parent_pop, false); end;

for i = 2:ngen
    fprintf('gen = %d\n', i)
    child_pop = selection(parent_pop, child_pop);
    child_pop = mutation_pop(child_pop);
    child_pop(:,obj_col) = 0;
    child_pop = evaluate_pop(child_pop, obj_func);
    mixed_pop = merge_pop(parent_pop, child_pop);
    parent_pop = fill_nondominated_sort(mixed_pop);
    
    % plot the current pareto front
    % if(do_plot) 
    %   show_plot(i, parent_pop, false, [1 2 3], [], ...
    %       [0.0, 1.0], [0.0, 1.0]); 
    % end;
    if(do_plot); show_plot(i, parent_pop, false, [1 2 3]); end;
    
    % save the current pop
    if(do_save); save_pop(i, parent_pop, false); end;
end
toc;
fprintf('Generations finished, now reporting solutions\n');

if(do_save)
    % save the final pop
    save_pop(i, parent_pop, false, 'final');
    % save the best pop
    save_pop(i, parent_pop, false, 'best');
end
fprintf('Routine successfully exited\n');