遗传算法(Python)

目录

1、概述

2、遗传算法易懂代码

(1)代码

(2)结果

3、遗传算法带约束 

(1)代码

(2)结果

4、参考文献(代码) 


1、概述

遗传算法的知识点已经梳理完了,现在直接上代码:

2、遗传算法易懂代码

(1)代码

#遗传算法:
import numpy as np
import matplotlib.pyplot as plt

def fitness(x):
    return x+16*np.sin(5*x)+10*np.cos(4*x)
class individual:
    def __init__(self):
        self.x=0
        self.fitness=0
    def __eq__(self,other):
        self.x=other.x
        self.fitness=other.fitness
def initPopulation(POP,N):
    for i in range(N):
        ind=individual()
        ind.x=np.random.uniform(-10,10)
        ind.fitness=fitness(ind.x)
        POP.append(ind)
def selection(N):
    return np.random.choice(N,2)
def crossover(parent1,parent2):
    child1,child2=individual(),individual()
    child1.x=0.9*parent1.x+0.1*parent2.x
    child2.x=0.1*parent2.x+0.9*parent2.x
    child1.fitness=fitness(child1.x)
    child2.fitness=fitness(child2.x)
    return child1,child2
def mutation(POP):
    ind=np.random.choice(POP)
    ind.x=np.random.uniform(-10,10)
    ind.fitness=fitness(ind.x)
def implement():
    N=40
    iter_N=400
    POP=[]
    initPopulation(POP,N)
    for it in range(iter_N):
        a,b=selection(N)
        if np.random.random()<0.65:
            child1,child2=crossover(POP[a],POP[b])
            new=sorted([POP[a],POP[b],child1,child2],key=lambda ind:ind.fitness,reverse=True)
            POP[a],POP[b]=new[0],new[1]
        if np.random.random()<0.1:
            mutation(POP)
    return POP
if __name__ =='__main__':
    pop=implement()

def func(x):
    return x+16*np.sin(5*x)+10*np.cos(4*x)
x=np.linspace(-10,10,100000)
y=func(x)
scatter_x=np.array([ind.x for ind in pop])
scatter_y=np.array([ind.fitness for ind in pop])
best=sorted(pop,key=lambda pop: pop.fitness,reverse=True)[0]
print('best_y',best.x)
print('best_y',best.fitness)
plt.plot(x,y)
plt.scatter(best.x,best.fitness,c='g',label='best point')
plt.legend()
plt.show()

(2)结果

                       遗传算法(Python)_第1张图片

 

3、遗传算法带约束 

(1)代码

import numpy as np
import matplotlib.pyplot as plt
from pylab import *
mpl.rcParams['font.sans-serif'] = ['SimHei']
mpl.rcParams['axes.unicode_minus'] = False

NP=50                
L=2                
Pc=0.5                 
Pm=0.1                
G=100                
Xmax=2                
Xmin=1                 
Ymax=0                
Ymin=-1               



def calc_f(X):
    a = 10
    pi = np.pi
    x = X[:, 0]
    y = X[:, 1]
    return 2 * a + x ** 2 - a * np.cos(2 * pi * x) + y ** 2 - a * np.cos(2 * 3.14 * y)
def calc_e(X):
    sumcost=[]

    for i in range(X.shape[0]):
        ee = 0
        """计算第一个约束的惩罚项"""
        e1 = X[i, 0] + X[i, 1] - 6
        ee += max(0, e1)
        """计算第二个约束的惩罚项"""
        e2 = 3 * X[i, 0] - 2 * X[i, 1] - 5
        ee += max(0, e2)
        sumcost.append(ee)
    return sumcost


##############遗传操作方法#########

def select(X, fitness):
    """根据轮盘赌法选择优秀个体"""
    fitness = 1 / fitness  # fitness越小表示越优秀,被选中的概率越大,做 1/fitness 处理
    fitness = fitness / fitness.sum()  # 归一化
    idx = np.array(list(range(X.shape[0])))
    X2_idx = np.random.choice(idx, size=X.shape[0], p=fitness)  # 根据概率选择
    X2 = X[X2_idx, :]
    return X2

def crossover(X, c):
    """按顺序选择2个个体以概率c进行交叉操作"""
    for i in range(0, X.shape[0], 2):
        parent1=X[i].copy() #父亲
        parent2=X[i + 1].copy()#母亲
        # 产生0-1区间的均匀分布随机数,判断是否需要进行交叉替换
        if np.random.rand() <= c:
            child1=(1-c)*parent1+c*parent2 #这是实数编码 的交叉形式 shape(2,)
            #child1=child1.reshape(-1,2)

            child2=c*parent1+(1-c)*parent2 #shape(2,)
            #child2=child2.reshape(1,2)
            #判断个体是否越限
            if child1[0]>Xmax or child1[0] < Xmin:
                child1[0]=np.random.uniform(Xmin, Xmax)
            if child1[1] > Ymax or child1[1]  Xmax or child2[0] < Xmin:
                child2[0] = np.random.uniform(Xmin, Xmax)
            if child2[1] > Ymax or child2[1] < Ymin:
                child2[1] = np.random.uniform(Ymin, Ymax)
            ######通过比较父辈和子代的适应度值和惩罚项 来决定要不要孩子
            X[i, :]=child1
            X[i + 1, :]=child2
    return X


def mutation(X, m):
    """变异操作"""
    for i in range(X.shape[0]):#遍历每一个个体
        # 产生0-1区间的均匀分布随机数,判断是否需要进行变异
        parent=X[i].copy()#父辈
        if np.random.rand() <= m:
                child = np.random.uniform(-1,2,(1,2))# 用随机赋值的方式进行变异 得到子代
                # 判断个体是否越限
                if child[:,0] > Xmax or child[:,0] < Xmin:
                    child[:,0] = np.random.uniform(Xmin, Xmax)
                if child[:,1] > Ymax or child[:,1] < Ymin:
                    child[:,1] = np.random.uniform(Ymin, Ymax)
                ######通过比较父辈和子代的适应度值和惩罚项 来决定要不要孩子
                X[i]=child
    return X

#子代和父辈之间的选择操作
def update_best(parent,parent_fitness,parent_e,child,child_fitness,child_e):
    """
        判
        :param parent: 父辈个体
        :param parent_fitness:父辈适应度值
        :param parent_e    :父辈惩罚项
        :param child:  子代个体
        :param child_fitness 子代适应度值
        :param child_e  :子代惩罚项

        :return: 父辈 和子代中较优者、适应度、惩罚项

        """
    # 规则1,如果 parent 和 child 都没有违反约束,则取适应度小的
    if parent_e <= 0.0000001 and child_e <= 0.0000001:
        if parent_fitness <= child_fitness:
            return parent,parent_fitness,parent_e
        else:
            return child,child_fitness,child_e
    # 规则2,如果child违反约束而parent没有违反约束,则取parent
    if parent_e < 0.0000001 and child_e  >= 0.0000001:
        return parent,parent_fitness,parent_e
    # 规则3,如果parent违反约束而child没有违反约束,则取child
    if parent_e >= 0.0000001 and child_e < 0.0000001:
        return child,child_fitness,child_e
    # 规则4,如果两个都违反约束,则取适应度值小的
    if parent_fitness <= child_fitness:
        return parent,parent_fitness,parent_e
    else:
        return child,child_fitness,child_e


def ga():
    """遗传算法主函数"""
    best_fitness = []  # 记录每次迭代的效果
    best_xy = []#存放最优xy
    f = np.random.uniform(-1, 2, (NP, 2))  # 初始化种群 (生成-1,2之间的随机数)shape (NP,2)
    for i in range(G):#遍历每一次迭代
        fitness=np.zeros((NP, 1))#存放适应度值
        ee=np.zeros((NP, 1)) #存放惩罚项值

        parentfit = calc_f(f)#计算父辈目标函数值
        parentee = calc_e(f)#计算父辈惩罚项
        parentfitness = parentfit + parentee #计算父辈适应度值   适应度值=目标函数值+惩罚项
        X2 = select(f, parentfitness)#选择
        X3 = crossover(X2, Pc)#交叉
        X4 = mutation(X3, Pm)#变异

        childfit = calc_f(X4)  # 子代目标函数值
        childee = calc_e(X4)  # 子代惩罚项
        childfitness = childfit + childee  # 子代适应度值

        for j in range(NP):#遍历每一个个体
            X4[j],fitness[j],ee[j] = update_best(f[j], parentfitness[j], parentee[j], X4[j], childfitness[j],childee[j])

        best_fitness.append(fitness.min())
        x, y = X4[fitness.argmin()]
        best_xy.append((x, y))
        f=X4
        # 多次迭代后的最终效果
    print("最优值是:%.5f" % best_fitness[-1])
    print("最优解是:x=%.5f, y=%.5f" % best_xy[-1])
    # 打印效果
    plt.plot(best_fitness, color='r')
    plt.show()
ga()

(2)结果

            遗传算法(Python)_第2张图片

4、参考文献(代码) 

    
————————————————
https://blog.csdn.net/kobeyu652453/article/details/109527260

你可能感兴趣的:(智能优化算法,全文检索,android,微信小程序,python)