遗传算法-求函数最优解

例题1：求目标函数Max{1-x²},-1<=x<=1，要求二进制编码，精确度为10^-3。

遗传算法的实现，流程如下：
    1.初始化种群：随机生成一定数量的染色体，每个染色体由一定数量的基因组成，每个基因的值为0或1。
    2.评估种群：对于每个染色体，计算其适应度，即目标函数的值。
    3.选择：根据染色体的适应度，选择一定数量的染色体作为下一代的父代。
    4.交叉：对于每一对父代，以一定的概率进行交叉操作，生成一个新的子代。
    5.变异：对于每个子代，以一定的概率进行变异操作，改变其中的一个或多个基因的值。
    6.生成下一代种群：将父代和子代合并，得到下一代种群。
    重复2-6步，直到达到停止条件（例如达到最大迭代次数或找到满足要求的解）。
 

解：在本题中，目标函数为f(x) = Max{1-x²},-1<=x<=1，其中x为染色体所代表的实数值。每个基因的值为0或1，表示实数值在二进制下的某一位。染色体的适应度为目标函数的值即f(x)。在每一代中，选择操作使用了轮盘赌选择，交叉操作使用了单点交叉，变异操作使用了单点变异。

参数配置

• POPULATION_SIZE：种群大小，即每一代中包含的个体数量，这里设置为100。
• CHROMOSOME_LENGTH：染色体长度，即每个个体的基因数量，这里设置为10。
• MUTATION_RATE：变异率，即每个基因发生变异的概率，这里设置为0.1。
• CROSSOVER_RATE：交叉率，即两个个体进行交叉的概率，这里设置为0.9。
• MAX_GENERATIONS：最大迭代次数，即算法运行的最大代数，这里设置为1000。
• PRECISION：精度，即当个体的适应度达到1.0时，算法停止运行，这里设置为0.001。

初始化种群

void initialize_population(vector<Chromosome>& population) {
    for (int i = 0; i < POPULATION_SIZE; i++) {
        Chromosome chromosome;
        for (int j = 0; j < CHROMOSOME_LENGTH; j++) {
            chromosome.genes.push_back(dis2(gen) < 0.5 ? 0 : 1);
        }
        population.push_back(chromosome);
    }
}

评估种群

double decode(vector<int> genes) {
    double x = 0.0;
    for (int i = 0; i < CHROMOSOME_LENGTH; i++) {
        x += genes[i] * pow(2, CHROMOSOME_LENGTH - i - 1);
    }
    return -1.0 + 2.0 * x / (pow(2, CHROMOSOME_LENGTH) - 1);
}
 
double fitness(vector<int> genes) {
    double x = decode(genes);
    return 1.0 - x * x;
}
 
void evaluate(vector<Chromosome>& population) {
    for (int i = 0; i < POPULATION_SIZE; i++) {
        population[i].fitness = fitness(population[i].genes);
    }
}

选择

void sort_population(vector<Chromosome>& population) {
    sort(population.begin(), population.end(), compare);
}
 
bool compare(Chromosome a, Chromosome b) {
    return a.fitness > b.fitness;
}

交叉

vector<int> crossover(vector<int> parent1, vector<int> parent2) {
    vector<int> child;
    for (int i = 0; i < CHROMOSOME_LENGTH; i++) {
        child.push_back(dis2(gen) < CROSSOVER_RATE ? parent1[i] : parent2[i]);
    }
    return child;
}

变异

void mutate(vector<int>& genes) {
    for (int i = 0; i < CHROMOSOME_LENGTH; i++) {
        if (dis2(gen) < MUTATION_RATE) {
            genes[i] = 1 - genes[i];
        }
    }
}

生成下一代种群

void next_generation(vector<Chromosome>& population) {
    vector<Chromosome> new_population;
    for (int i = 0; i < POPULATION_SIZE; i++) {
        vector<int> parent1_genes = population[i].genes;
        vector<int> parent2_genes = population[dis(gen)].genes;
        vector<int> child_genes = crossover(parent1_genes, parent2_genes);
        mutate(child_genes);
        Chromosome child;
        child.genes = child_genes;
        new_population.push_back(child);
    }
    evaluate(new_population);
    sort_population(new_population);
    population = new_population;
}

停止条件

for (int i = 0; i < MAX_GENERATIONS; i++) {
    if (population[0].fitness >= 1.0 - PRECISION) {
        break;
    }
    next_generation(population);
}

程序结果

最优解：x = 0.000977517,f(x) = 0.999999

完整代码

#include <iostream>
#include <cmath>
#include <vector>
#include <algorithm>
#include <random>
using namespace std;
 
const int POPULATION_SIZE = 100;
const int CHROMOSOME_LENGTH = 10;
const double MUTATION_RATE = 0.1;
const double CROSSOVER_RATE = 0.9;
const int MAX_GENERATIONS = 1000;
const double PRECISION = 0.001;
 
random_device rd;
mt19937 gen(rd());
uniform_real_distribution<> dis(-1.0, 1.0);
uniform_real_distribution<> dis2(0.0, 1.0);
 
struct Chromosome {
    vector<int> genes;
    double fitness;
};
 
double decode(vector<int> genes) {
    double x = 0.0;
    for (int i = 0; i < CHROMOSOME_LENGTH; i++) {
        x += genes[i] * pow(2, CHROMOSOME_LENGTH - i - 1);
    }
    return -1.0 + 2.0 * x / (pow(2, CHROMOSOME_LENGTH) - 1);
}
 
double fitness(vector<int> genes) {
    double x = decode(genes);
    return 1.0 - x * x;
}
 
void evaluate(vector<Chromosome>& population) {
    for (int i = 0; i < POPULATION_SIZE; i++) {
        population[i].fitness = fitness(population[i].genes);
    }
}
 
bool compare(Chromosome a, Chromosome b) {
    return a.fitness > b.fitness;
}
 
void sort_population(vector<Chromosome>& population) {
    sort(population.begin(), population.end(), compare);
}
 
void initialize_population(vector<Chromosome>& population) {
    for (int i = 0; i < POPULATION_SIZE; i++) {
        Chromosome chromosome;
        for (int j = 0; j < CHROMOSOME_LENGTH; j++) {
            chromosome.genes.push_back(dis2(gen) < 0.5 ? 0 : 1);
        }
        population.push_back(chromosome);
    }
}
 
vector<int> crossover(vector<int> parent1, vector<int> parent2) {
    vector<int> child;
    for (int i = 0; i < CHROMOSOME_LENGTH; i++) {
        child.push_back(dis2(gen) < CROSSOVER_RATE ? parent1[i] : parent2[i]);
    }
    return child;
}
void mutate(vector<int>& genes) {
    for (int i = 0; i < CHROMOSOME_LENGTH; i++) {
        if (dis2(gen) < MUTATION_RATE) {
            genes[i] = 1 - genes[i];
        }
    }
}
 
void next_generation(vector<Chromosome>& population) {
    vector<Chromosome> new_population;
    for (int i = 0; i < POPULATION_SIZE; i++) {
        vector<int> parent1_genes = population[i].genes;
        vector<int> parent2_genes = population[dis(gen)].genes;
        vector<int> child_genes = crossover(parent1_genes, parent2_genes);
        mutate(child_genes);
        Chromosome child;
        child.genes = child_genes;
        new_population.push_back(child);
    }
    evaluate(new_population);
    sort_population(new_population);
    population = new_population;
}
 
int main() {
    vector<Chromosome> population;
    initialize_population(population);
    evaluate(population);
    sort_population(population);
    for (int i = 0; i < MAX_GENERATIONS; i++) {
        if (population[0].fitness >= 1.0 - PRECISION) {
            break;
        }
        next_generation(population);
    }
    cout << "最优解: x = " << decode(population[0].genes) << ", f(x) = " << population[0].fitness << endl;
    return 0;
}