MasterMind/main (conflicted copy 2020-02-03 100957).cpp

#include <iostream>
#include <cstdlib>
#include <ctime>
#include <numeric>
#include <algorithm>
#include <vector>
#include <random>
#include <chrono>
#include <cmath>
#include <windows.h>

#include <limits>
#include <fstream>
#include <sstream>
#include <stdlib.h>

#include "EA.h"
#include "RLS.h"
#include "analysis.h"
#include "PEA.h"

#include "ea-complexity.h"
#include "crossover-rate.h"
#include "scripts.h"
#include "scripts2.h"
#include "bestPGA.h"
#include "bestPGA.h"
#include "bestPEA.h"
#include "algo-complexity.h"

using namespace std;

struct Data {
    Data() : nb_goods{0} {};

    double r;
    double s;
    double t;
    double u;

    double a;
    double b;
    double c;
    double d;

    double alpha;
    double beta;
    double gamma;
    double lambda;

    int nb_goods;
};

int main(int argc, char** argv) {
    srand(time(NULL));
    std::default_random_engine randomizer{static_cast<long unsigned int>(time(0))};

    if(argc >= 2) {
        int num_experiment = atoi(argv[1]);
        switch(num_experiment) {
            case 1:
                cout << "Experiment: Optimal Crossover probability" << endl;
                switch(atoi(argv[2])) {
                    case 0: // Start from beginning
                        script_crossover_proba_global_optim(randomizer, "global-crossover-proba.csv", false, 5, 5);
                        break;
                    case 1: // Restart from a specific point
                        {
                            stringstream ss;
                            ss << "global-crossover-proba-" << atoi(argv[3]) << "-" << atoi(argv[4]) << ".csv";
                            script_crossover_proba_global_optim(randomizer, ss.str(), true, atoi(argv[3]), atoi(argv[4]));
                        }break;
                    case 2: // Filtering
                        crossover_model_filtering(argv[3], argv[4], argv[5], atoi(argv[6]), atoi(argv[7]));
                        break;
                    case 3: // Local optimization of model
                        crossover_model_optimization(argv[3]);
                        break;
                    case 4: // Validation of model
                        crossover_model_validation(argv[3], atof(argv[4]), atof(argv[5]), atof(argv[6]), atof(argv[7]));
                        break;
                }
                return 0;

            case 2:
                cout << "Experiment: EA Complexity" << endl;
                switch(atoi(argv[2])) {
                    case 1:
                        {
                            int n_initial = (argc>=4) ? atoi(argv[3]) : 3;
                            stringstream ss;
                            ss << "complexity-ea-n-" << n_initial << ".csv";
                            ofstream outfile(ss.str());
                            for(int n=n_initial; n<30; n++) {
                                MasterMind mm(n, n, randomizer);
                                for(int mu=1; mu<50; mu+=2) {
                                    cout << " n=" << n << ": mu=" << mu << endl;
                                    for(int lambda=1; lambda<50; lambda+=2) {
                                        EA algo(mu, lambda, PLUS, 1./n, randomizer);
                                        algo.set_problem(&mm);
                                        Statistic stat = avg_calls(&algo, 1000, -1, 1.9623414611);
                                        outfile << n << ";" << mu << ";" << lambda << ";" << stat.average << ";" << stat.alea << endl;
                                    }
                                }
                            }
                        }
                        break;
                    case 2: // Filtering
                        ea_complexity_filtering(argv[3], argv[4], atoi(argv[5]), atoi(argv[6]), atoi(argv[7]));
                        break;
                }
                return 0;
        }
    }
    ////////////////////////////////////////////////////

    Data datas[30];

    std::ifstream infile("complexity-ea.csv");

    std::string line;
    while(getline(infile, line)) {
        int n, mu, lda;
        double avg, alea;
        sscanf(line.c_str(), "%d;%d;%d;%lf;%lf", &n, &mu, &lda, &avg, &alea);

        if((mu==5 || mu==45) && (lda==5 || lda==45)) {
            if(mu==5) {
                if(lda==5)
                    datas[n].r = avg;
                else
                    datas[n].s = avg;
            } else {
                if(lda==5)
                    datas[n].t = avg;
                else
                    datas[n].u = avg;
            }
        }
    }

    for(int n=5; n<30; n++) {
        Data dat = datas[n];
        datas[n].a = (dat.s - dat.r) / (45. - 5.);
        datas[n].b = dat.s - 45.*datas[n].a;
        datas[n].c = (dat.u - dat.t) / (45. - 5.);
        datas[n].b = dat.u - 45.*datas[n].c;
    }
    for(int n=5; n<30; n++) {
        Data dat = datas[n];
        datas[n].alpha = (dat.c - dat.a) / (45. - 5.);
        datas[n].gamma = dat.a - 5*datas[n].alpha;
        datas[n].beta = (dat.d - dat.b) / (45. - 5.);
        datas[n].lambda = dat.b - 5*datas[n].beta;
    }

    std::ifstream in2file("complexity-ea.csv");
    while(getline(in2file, line)) {
        int n, mu, lda;
        double avg, alea;
        sscanf(line.c_str(), "%d;%d;%d;%lf;%lf", &n, &mu, &lda, &avg, &alea);

        Data d = datas[n];
        double prev = d.alpha*mu*lda + d.beta*mu + d.gamma*lda + d.lambda;

        if(abs(avg-prev) <= alea)
            datas[n].nb_goods++;
    }
    for(int n=5; n<30; n++) {
        Data d = datas[n];
        cout << n << ": " << d.alpha << "*mu*lda + " << d.beta << "*mu + " << d.gamma << "*lda + " << d.lambda << " => " << 100*d.nb_goods/625. << "%" << endl;
    }

    return 0;


/*    RLS algo;
    vector<Statistic> stats = compute_complexity(randomizer, &algo, 10, 1000, 10, 100, 1.9842169516, true);
    save_in_file("RLS-complexity", stats);
    return 0;
*/
    //script_optimal_proba(randomizer);
/*
    pea_optimisation(10, 20, randomizer);
    return 0;
*/

    int nb_colors = 20;
    int nb_cases = 20;
    int n=20;
/*
    MasterMind mm(nb_colors, nb_cases, randomizer);
    RLS algo1;
    EA algo2(1, 1, PLUS, 1./n, randomizer);
    EA algo3(2, 1, PLUS, 1./n, randomizer);
    EA algo4(4, 1, PLUS, 1./n, randomizer);
    EA algo5(8, 1, PLUS, 1./n, randomizer);
    EA algo6(16, 1, PLUS, 1./n, randomizer);
    algo1.set_problem(&mm);
    algo2.set_problem(&mm);
    algo3.set_problem(&mm);
    algo4.set_problem(&mm);
    algo5.set_problem(&mm);
    algo6.set_problem(&mm);
    vector<SearchHeuristic*> algos = {&algo1, &algo2, &algo3, &algo4, &algo5, &algo6};
*/
    MasterMind mm(nb_colors, nb_cases, randomizer);
    RLS algo1;
    EA algo2(1, 1, PLUS, 1./n, randomizer);
    PEA algo3(1, 1, PLUS, {}, randomizer);
    algo1.set_problem(&mm);
    algo2.set_problem(&mm);
    algo3.set_problem(&mm);
    vector<SearchHeuristic*> algos = {&algo1, &algo2, &algo3};

    Statistic stat = avg_calls(&algo1, 20, -1, 3.883);
    int max_nb_calls = ceil(stat.average+stat.alea)*2;
    cmp_evolution(algos, mm, max_nb_calls, 100);
    return 0;
/*
    Statistic* curves[6];
    Statistic* dcurves[6];
    int length = 0;
    for(int i=0; i<algos.size(); i++) {
        curves[i] = avg_evolution(algos[i], max_nb_calls, 20);
        dcurves[i] = derivate(curves[i], max_nb_calls);
        length = max_nb_calls-1;
        for(int k=0; k<7; k++) {
            Statistic* cdcurve = compress(dcurves[i], length);
            delete[] dcurves[i];
            dcurves[i] = cdcurve;
            length = ceil(length/2.)-1;
        }
        for(int k=0; k<15; k++) {
            Statistic* sdcurve = smooth(dcurves[i], length);
            delete[] dcurves[i];
            dcurves[i] = sdcurve;
        }
    }

    ofstream testfile("42.csv");
    for(int i=0; i<length; i++) {
        for(int j=0; j<algos.size(); j++)
            testfile << dcurves[j][i].average << ";";
        testfile << endl;
    }
    ofstream testfile2("43.csv");
    for(int i=0; i<max_nb_calls; i++) {
        for(int j=0; j<algos.size(); j++)
            testfile2 << curves[j][i].average << ";";
        testfile2 << endl;
    }
    for(int j=0; j<algos.size(); j++) {
        delete[] dcurves[j];
        delete[] curves[j];
    }

    return 0;

    int initial_n = 3;
    int final_n = 25;

    /// Mu-Lambda Optimisation
    MuLambdaOptimizer mlo(initial_n, randomizer);
    mlo.set_nb_contexts(final_n - initial_n);
    mlo.set_study_interval(1, 100, 4);
    mlo.add_test_wave(100, 2.4751374205, 0);
    mlo.add_test_wave(1000, 2.4365918443, 1);
    mlo.add_test_wave(10000, 2.4327997145, 2);
    //mlo.run();


/*
    //optimal_mulambda(3, 16, randomizer, 1, 50, 1, 100, 1.9842169516);
    return 0;

    optimal_lambda(20, 1, 2, randomizer, 1, 100, 1, 50, 1.9842169516, true);
    return 0;
*/
/*    int nb_tests = 100;
    MasterMind mm(nb_colors, nb_cases, randomizer);
    RLS algo1;
    GA algo3(10, 10, PLUS, 1./nb_cases, 2./nb_cases, UNIFORM_CROSSOVER, GA_V1, randomizer);

    algo1.set_problem(&mm);
    Statistic stat = avg_calls(&algo1, 20, -1, 3.883);
    int max_nb_calls = ceil(stat.average);

    cmp_evolution({&algo1, &algo3}, mm, max_nb_calls, nb_tests);
    return 0;
*/
    //cout << "=========================" << endl;
    //optimal_proba(5,15, randomizer);
    //optimal_proba_v2(5,15, randomizer);
    //optimal_proba_vmath2_rec(3, 16, randomizer, 0.05, 0.5, 0.01, {20, 200, 2000}, {2.3362421597,2.3451370823,3.2953981367});
    //optimal_proba_vmath2_rec(16, 25, randomizer, 0.02, 0.2, 0.005, {20, 200, 2000}, {2.3362421597,2.3451370823,3.2953981367});

    cout << "FINISHED !" << endl;
    std::cin.ignore(std::numeric_limits<std::streamsize>::max(), '\n');
    std::cin.get();

    return 0;

    cout << "=========================" << endl;
    for(int n=5; n<6; n++) {
        vector<double> avg_list;
        vector<double> lambdas;

        int lambda_init = 1;
        int lambda_final = 20;
        int nb_tests = 10000;

        auto start = start_chrono();
        for(int lambda=lambda_init; lambda<lambda_final; lambda++) {
            EA algo(1, lambda, PLUS, 1./n, randomizer);
            MasterMind mm(n, n, randomizer);
            algo.set_problem(&mm);
            Statistic stat = avg_calls(&algo, nb_tests);
            double avg = stat.average;
            avg_list.push_back(avg);
            lambdas.push_back(lambda);
            cout << " -> lambda=" << lambda << ", nbcalls=" << avg << endl;
        }
        auto stop = stop_chrono();
        int elapsed_milliseconds = interval_chrono(start, stop);

        int min_ind = 0;
        for(unsigned int i=1; i<avg_list.size(); i++)
            if(avg_list[min_ind] > avg_list[i])
                min_ind = i;
        double lambda_best = lambdas[min_ind];
        cout << "n=" << n << ": " << lambda_best << " in " << elapsed_milliseconds/1000. << " seconds " << "(" << nb_tests*(lambda_final-lambda_init) << " problems)" << endl;
    }

    return 0;
}