用遗传算法求解香蕉函数最小值的python代码

以下是使用遗传算法求解香蕉函数最小值的 Python 代码示例：

import random

# 定义香蕉函数
def banana_function(x, y):
    return (1 - x)**2 + 100 * (y - x**2)**2

# 定义个体类
class Individual:
    def __init__(self, x, y):
        self.x = x
        self.y = y
        self.fitness = banana_function(x, y)

# 定义遗传算法类
class GeneticAlgorithm:
    def __init__(self, population_size, mutation_rate, crossover_rate, elitism_rate, tournament_size):
        self.population_size = population_size
        self.mutation_rate = mutation_rate
        self.crossover_rate = crossover_rate
        self.elitism_rate = elitism_rate
        self.tournament_size = tournament_size

    # 初始化种群
    def initialize_population(self):
        population = []
        for i in range(self.population_size):
            x = random.uniform(-10, 10)
            y = random.uniform(-10, 10)
            individual = Individual(x, y)
            population.append(individual)
        return population

    # 选择操作
    def selection(self, population):
        selected_population = []
        for i in range(self.population_size):
            tournament = random.sample(population, self.tournament_size)
            winner = min(tournament, key=lambda x: x.fitness)
            selected_population.append(winner)
        return selected_population

    # 交叉操作
    def crossover(self, population):
        offspring_population = []
        for i in range(int(self.population_size * self.crossover_rate)):
            parent1, parent2 = random.sample(population, 2)
            offspring_x = (parent1.x + parent2.x) / 2
            offspring_y = (parent1.y + parent2.y) / 2
            offspring = Individual(offspring_x, offspring_y)
            offspring_population.append(offspring)
        return offspring_population

    # 变异操作
    def mutation(self, population):
        mutated_population = []
        for i in range(int(self.population_size * self.mutation_rate)):
            individual = random.choice(population)
            mutated_x = individual.x + random.uniform(-0.1, 0.1)
            mutated_y = individual.y + random.uniform(-0.1, 0.1)
            mutated_individual = Individual(mutated_x, mutated_y)
            mutated_population.append(mutated_individual)
        return mutated_population

    # 精英保留操作
    def elitism(self, population):
        sorted_population = sorted(population, key=lambda x: x.fitness)
        elitism_size = int(self.population_size * self.elitism_rate)
        elitism_population = sorted_population[:elitism_size]
        return elitism_population

    # 进化操作
    def evolve(self, population):
        selected_population = self.selection(population)
        offspring_population = self.crossover(selected_population)
        mutated_population = self.mutation(offspring_population)
        new_population = self.elitism(population) + selected_population + mutated_population
        return new_population

    # 求解函数最小值
    def solve(self):
        population = self.initialize_population()
        best_individual = min(population, key=lambda x: x.fitness)
        for i in range(100):
            population = self.evolve(population)
            current_best_individual = min(population, key=lambda x: x.fitness)
            if current_best_individual.fitness < best_individual.fitness:
                best_individual = current_best_individual
        return best_individual.x, best_individual.y

# 测试代码
if __name__ == '__main__':
    ga = GeneticAlgorithm(population_size=100, mutation_rate=0.2, crossover_rate=0.5, elitism_rate=0.1, tournament_size=5)
    x, y = ga.solve()
    print(f"The minimum value of the banana function is {banana_function(x, y)} at ({x}, {y})")

在这个示例中，我们使用了一个包含 100 个个体的种群，进行了 100 次迭代，每次迭代使用了 5 轮锦标赛选择，以及 20% 的个体进行变异，50% 的个体进行交叉，和 10% 的精英保留策略。最后输出找到的香蕉函数最小值和对应的 $(x, y)$ 坐标。