遗传算法之: One Max Problem

已于 2023-03-27 15:04:02 修改
阅读量2k 收藏 9
点赞数 5
分类专栏： GeneticAlgorithm 文章标签： 遗传算法 OneMaxProblem 算法
于 2020-09-14 17:41:21 首次发布
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/u014281392/article/details/108583654
版权

遗传算法之: One Max Problem

OMP问题是一个最大化一个0,1比特串中1的数量的简单问题.

简单描述为: X → = { x 1 , x 2 , x 3 , . . . x N } \overrightarrow{X} = \{x_1, x_2, x_3, ... x_N\} X ={x1​,x2​,x3​,...xN​}, 且 x i ∈ { 0 , 1 } x_i \in \{0, 1\} xi​∈{0,1},目标是使下面等式最大化:
F ( X → ) = ∑ i = 1 N x i F(\overrightarrow{X}) = \sum_{i=1}^{N} x_i F(X )=i=1∑N​xi​

这个例子中, N = 100 N = 100 N=100

进化策略1:

  1. 随机生成父比特串
  2. 计算适应度
  3. 随机突变
  4. 计算突变生成子串的适应度
  5. 比较当前子串是否是当前最优
    • 是,继续在此基础上进行突变
    • 否,从父串突变
  6. 直到 F ( X → ) F(\overrightarrow{X}) F(X ) = 100结束

进化策略2:

  1. 随机生成父群
  2. 计算适应度,挑选父群中top n
  3. top n父群进行交叉变异
  4. 计算适应度,选择top n
  5. 循环3, 4
  6. 直到有子串满足结束条件 F ( X → ) = 100 F(\overrightarrow{X}) = 100 F(X )=100

进化策略3:

  1. 在策略2的基础上,交叉后增加,随机突变,丰富交叉样本的多样性

1. 随机突变

import random
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline

def generate_sigle_parent(length, geneSet):
    parentGene = np.array([np.random.choice(geneSet) for i in range(length)])
    return parentGene

def get_fitness(gene):
    return np.sum(gene)

def display(gene):
    geneStr = ''.join(map(str, gene))
    fitness = get_fitness(gene)
    print('gene: %s, fitness: %d'%(geneStr, fitness))

def unisexual_mutate(parentGene, geneSet):
    childGene = parentGene.copy()
    index = random.randrange(0, len(parentGene))
    newGene, alternate = random.sample(geneSet, 2)
    if childGene[index] != newGene:
        childGene[index] = newGene
    else:
        childGene[index] = alternate
    return childGene

def evolution_v1(geneSet, target=100):
    gene_list, fitness_list = [], []
    bestGene = generate_sigle_parent(target,geneSet)
    bestFitness = get_fitness(bestGene)
    gene_list.append(bestGene)
    fitness_list.append(bestFitness)
    display(bestGene)
    print('--------first generation---------')
    if bestFitness >= target:
        return bestGene
    while True:
        childGene = unisexual_mutate(bestGene, geneSet)
        fitness = get_fitness(childGene)
        gene_list.append(childGene) # collect
        fitness_list.append(fitness) # collect
        if fitness <= bestFitness:
            continue
        display(childGene)
        if fitness >= target:
            return childGene, gene_list, fitness_list
        bestGene = childGene
        bestFitness = fitness
单性遗传(随机突变)
best_gene, gene_log1, fitness_log1 = evolution_v1(geneSet=[0, 1], target=100)

gene: 1100011100001110000011011110110100001100100101100001111010011111100001101100001011000001111110000110, fitness: 49
--------first generation---------
gene: 1100011100001110000011011110110100001100100101100001111010011111100001101100001011010001111110000110, fitness: 50
gene: 1100011100001110000011011110110100001100100101101001111010011111100001101100001011010001111110000110, fitness: 51
gene: 1100011100001110000011011110110100001100100101101001111010011111110001101100001011010001111110000110, fitness: 52
gene: 1100011100001110000011011110110100001100100101101001111010011111110001101100001011010001111110100110, fitness: 53
gene: 1100011100001110000011011110110100001100100101101001111010011111110001101110001011010001111110100110, fitness: 54
gene: 1100011100001110000011011110110100001100100101101001111010011111110001101110101011010001111110100110, fitness: 55
gene: 1100011100001110010011011110110100001100100101101001111010011111110001101110101011010001111110100110, fitness: 56
gene: 1100011110001110010011011110110100001100100101101001111010011111110001101110101011010001111110100110, fitness: 57
gene: 1100011110001110110011011110110100001100100101101001111010011111110001101110101011010001111110100110, fitness: 58
gene: 1100011110001110110011011110110101001100100101101001111010011111110001101110101011010001111110100110, fitness: 59
gene: 1100011110001110110011011110110101001100100101101001111010011111110001101110101011010001111110100111, fitness: 60
gene: 1100011110011110110011011110110101001100100101101001111010011111110001101110101011010001111110100111, fitness: 61
gene: 1100011110011110110011011110111101001100100101101001111010011111110001101110101011010001111110100111, fitness: 62
gene: 1100011110011110111011011110111101001100100101101001111010011111110001101110101011010001111110100111, fitness: 63
gene: 1100011110011110111011011110111111001100100101101001111010011111110001101110101011010001111110100111, fitness: 64
gene: 1100011110011110111011011110111111001100100101101001111010011111110001111110101011010001111110100111, fitness: 65
gene: 1100011110011110111011011110111111011100100101101001111010011111110001111110101011010001111110100111, fitness: 66
gene: 1100011110011110111011111110111111011100100101101001111010011111110001111110101011010001111110100111, fitness: 67
gene: 1100011110011110111011111110111111011101100101101001111010011111110001111110101011010001111110100111, fitness: 68
gene: 1100011110011110111011111110111111011101100101111001111010011111110001111110101011010001111110100111, fitness: 69
gene: 1100011110011110111011111110111111011101110101111001111010011111110001111110101011010001111110100111, fitness: 70
gene: 1100011110011110111011111110111111011101111101111001111010011111110001111110101011010001111110100111, fitness: 71
gene: 1100011110011110111011111110111111011101111101111001111010011111110101111110101011010001111110100111, fitness: 72
gene: 1100011110011110111111111110111111011101111101111001111010011111110101111110101011010001111110100111, fitness: 73
gene: 1100011110011110111111111110111111011101111101111001111010011111110101111110101011010001111110101111, fitness: 74
gene: 1100011110011110111111111110111111011101111101111001111010011111110101111110101011010011111110101111, fitness: 75
gene: 1100011110011110111111111110111111011101111101111001111010011111110101111110101111010011111110101111, fitness: 76
gene: 1100011110011110111111111110111111111101111101111001111010011111110101111110101111010011111110101111, fitness: 77
gene: 1100011110011110111111111110111111111101111101111001111010011111111101111110101111010011111110101111, fitness: 78
gene: 1100011110011111111111111110111111111101111101111001111010011111111101111110101111010011111110101111, fitness: 79
gene: 1100011110011111111111111110111111111101111101111001111010011111111101111110101111010011111110111111, fitness: 80
gene: 1100111110011111111111111110111111111101111101111001111010011111111101111110101111010011111110111111, fitness: 81
gene: 1100111110011111111111111110111111111101111101111001111010011111111101111110101111010111111110111111, fitness: 82
gene: 1100111110011111111111111110111111111101111101111001111010011111111101111110101111110111111110111111, fitness: 83
gene: 1101111110011111111111111110111111111101111101111001111010011111111101111110101111110111111110111111, fitness: 84
gene: 1101111110011111111111111110111111111101111111111001111010011111111101111110101111110111111110111111, fitness: 85
gene: 1101111110011111111111111110111111111101111111111001111010011111111101111110111111110111111110111111, fitness: 86
gene: 1101111110011111111111111110111111111101111111111001111010011111111111111110111111110111111110111111, fitness: 87
gene: 1101111110011111111111111110111111111101111111111001111011011111111111111110111111110111111110111111, fitness: 88
gene: 1101111110011111111111111110111111111101111111111001111011111111111111111110111111110111111110111111, fitness: 89
gene: 1111111110011111111111111110111111111101111111111001111011111111111111111110111111110111111110111111, fitness: 90
gene: 1111111110011111111111111110111111111101111111111011111011111111111111111110111111110111111110111111, fitness: 91
gene: 1111111110011111111111111110111111111101111111111011111111111111111111111110111111110111111110111111, fitness: 92
gene: 1111111110011111111111111110111111111111111111111011111111111111111111111110111111110111111110111111, fitness: 93
gene: 1111111110011111111111111110111111111111111111111011111111111111111111111110111111110111111111111111, fitness: 94
gene: 1111111111011111111111111110111111111111111111111011111111111111111111111110111111110111111111111111, fitness: 95
gene: 1111111111011111111111111110111111111111111111111011111111111111111111111110111111111111111111111111, fitness: 96
gene: 1111111111011111111111111110111111111111111111111011111111111111111111111111111111111111111111111111, fitness: 97
gene: 1111111111111111111111111110111111111111111111111011111111111111111111111111111111111111111111111111, fitness: 98
gene: 1111111111111111111111111111111111111111111111111011111111111111111111111111111111111111111111111111, fitness: 99
gene: 1111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111, fitness: 100
进化过程

在这里插入图片描述

2.随机交叉

def generate_parents_group(length, geneSet, group_size=10):
    parentGeneGroup = [np.array([np.random.choice(geneSet) for i in range(length)]) for i in range(group_size)]
    return parentGeneGroup

贪心选择

def select_topN_gene(GeneGroup, topn=5):
    fitness_list =[-1*get_fitness(gene) for gene in GeneGroup]
    sorted_index = np.argsort(fitness_list)
    selected_Gene = [GeneGroup[i] for i in sorted_index[:topn]]
    return selected_Gene

import itertools
def bisexual_mutate(GeneGroup, alpha=0.3):   
    childGenes = []
    for couple in itertools.combinations(GeneGroup, 2):
        childGene = couple[0].copy()
        length = len(childGene)
        random_location = random.sample(list(np.arange(length)), int(length*alpha))
        childGene[random_location] = couple[1][random_location] # 交叉
        childGenes.append(childGene)
    return childGenes

def evolution_v2(geneSet, target=100, group_size=10, topN=5, alpha=0.5):
    gene_list = []
    mean_fitness = []
    max_fitness = []
    ParentGeneGroup = generate_parents_group(target, geneSet,group_size)
    #ParentGeneGroup = bisexual_mutate(ParentGeneGroup,alpha=0.2)
    selectedPGroup = select_topN_gene(ParentGeneGroup, topN)
    for gene in selectedPGroup:
        display(gene)
    print('Selected Gene Group')
    first_fitness = [get_fitness(gene) for gene in selectedPGroup]
    gene_list.append(selectedPGroup)
    mean_fitness.append(np.mean(first_fitness))
    max_fitness.append(np.max(first_fitness))
    # 
    for fitness in first_fitness:
        if fitness >= target:
            return selected_group
    #while True:
    for i in range(500):
        # 生成下一代
        childGeneGroup = bisexual_mutate(selectedPGroup,alpha=alpha)
        # 筛选
        selectedCGroup = select_topN_gene(childGeneGroup)
        gene_list.append(selectedCGroup) # collect
        # 上一代和下一代平均适应度进行比较
        mean_c_fitness = np.mean([get_fitness(gene) for gene in selectedCGroup])
        mean_p_fitness = np.mean([get_fitness(gene) for gene in selectedPGroup])
        max_c_fitness = np.max([get_fitness(gene) for gene in selectedCGroup])
        mean_fitness.append(mean_c_fitness)
        max_fitness.append(max_c_fitness)
        if mean_c_fitness <= mean_p_fitness:
            continue
        print('Selected Gene Mean Fitness :%f'%mean_c_fitness)
        if max_c_fitness >= target:
            return selectedCGroup, gene_list, mean_fitness
        selectedPGroup = selectedCGroup
    return selectedCGroup, gene_list, mean_fitness

#bestGroup, GeneGroup_list, Avg_fitness = evolution_v2(geneSet, alpha=0.1)  # alpha 交叉系数
result_1 = evolution_v2(geneSet, alpha=0.1)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :57.400000
Selected Gene Mean Fitness :59.600000
Selected Gene Mean Fitness :60.800000
Selected Gene Mean Fitness :62.600000
Selected Gene Mean Fitness :64.400000
Selected Gene Mean Fitness :67.000000
Selected Gene Mean Fitness :68.000000
Selected Gene Mean Fitness :68.400000
Selected Gene Mean Fitness :69.000000
Selected Gene Mean Fitness :69.200000
Selected Gene Mean Fitness :69.800000
Selected Gene Mean Fitness :70.000000

result_2 = evolution_v2(geneSet, alpha=0.2)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :56.000000
Selected Gene Mean Fitness :59.000000
Selected Gene Mean Fitness :62.400000
Selected Gene Mean Fitness :66.400000
Selected Gene Mean Fitness :69.400000
Selected Gene Mean Fitness :70.400000
Selected Gene Mean Fitness :71.000000
Selected Gene Mean Fitness :71.200000
Selected Gene Mean Fitness :71.800000
Selected Gene Mean Fitness :72.000000

result_3 = evolution_v2(geneSet, alpha=0.3)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :58.400000
Selected Gene Mean Fitness :62.600000
Selected Gene Mean Fitness :65.200000
Selected Gene Mean Fitness :67.200000
Selected Gene Mean Fitness :70.200000
Selected Gene Mean Fitness :71.000000
Selected Gene Mean Fitness :72.600000
Selected Gene Mean Fitness :73.000000
Selected Gene Mean Fitness :73.800000
Selected Gene Mean Fitness :74.000000

result_4 = evolution_v2(geneSet, alpha=0.4)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :55.600000
Selected Gene Mean Fitness :58.000000
Selected Gene Mean Fitness :59.200000
Selected Gene Mean Fitness :61.400000
Selected Gene Mean Fitness :65.400000
Selected Gene Mean Fitness :67.400000
Selected Gene Mean Fitness :69.800000
Selected Gene Mean Fitness :70.800000
Selected Gene Mean Fitness :71.200000
Selected Gene Mean Fitness :71.600000
Selected Gene Mean Fitness :72.000000

result_5 = evolution_v2(geneSet, alpha=0.5)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :54.600000
Selected Gene Mean Fitness :55.200000
Selected Gene Mean Fitness :58.600000
Selected Gene Mean Fitness :60.200000
Selected Gene Mean Fitness :61.600000
Selected Gene Mean Fitness :64.800000
Selected Gene Mean Fitness :65.600000
Selected Gene Mean Fitness :66.600000
Selected Gene Mean Fitness :66.800000
Selected Gene Mean Fitness :67.800000
Selected Gene Mean Fitness :68.400000
Selected Gene Mean Fitness :69.000000
Selected Gene Mean Fitness :69.200000
Selected Gene Mean Fitness :70.600000
Selected Gene Mean Fitness :70.800000
Selected Gene Mean Fitness :71.000000
Selected Gene Mean Fitness :71.200000
Selected Gene Mean Fitness :71.600000
Selected Gene Mean Fitness :72.000000

result_6 = evolution_v2(geneSet, alpha=0.6)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :56.000000
Selected Gene Mean Fitness :57.800000
Selected Gene Mean Fitness :59.400000
Selected Gene Mean Fitness :60.800000
Selected Gene Mean Fitness :62.000000
Selected Gene Mean Fitness :64.800000
Selected Gene Mean Fitness :66.000000
Selected Gene Mean Fitness :67.600000
Selected Gene Mean Fitness :69.200000
Selected Gene Mean Fitness :69.600000
Selected Gene Mean Fitness :71.000000
Selected Gene Mean Fitness :72.600000
Selected Gene Mean Fitness :72.800000
Selected Gene Mean Fitness :73.400000
Selected Gene Mean Fitness :73.800000
Selected Gene Mean Fitness :74.000000

result_7 = evolution_v2(geneSet, alpha=0.7)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :55.000000
Selected Gene Mean Fitness :57.600000
Selected Gene Mean Fitness :58.600000
Selected Gene Mean Fitness :59.600000
Selected Gene Mean Fitness :61.200000
Selected Gene Mean Fitness :63.600000
Selected Gene Mean Fitness :64.800000
Selected Gene Mean Fitness :66.600000
Selected Gene Mean Fitness :67.400000
Selected Gene Mean Fitness :68.400000
Selected Gene Mean Fitness :68.600000
Selected Gene Mean Fitness :69.000000
Selected Gene Mean Fitness :70.200000
Selected Gene Mean Fitness :71.200000
Selected Gene Mean Fitness :71.600000
Selected Gene Mean Fitness :72.200000
Selected Gene Mean Fitness :72.400000
Selected Gene Mean Fitness :72.800000
Selected Gene Mean Fitness :73.000000

result_8 = evolution_v2(geneSet, alpha=0.8)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :52.400000
Selected Gene Mean Fitness :54.600000
Selected Gene Mean Fitness :55.400000
Selected Gene Mean Fitness :56.200000
Selected Gene Mean Fitness :57.200000
Selected Gene Mean Fitness :58.800000
Selected Gene Mean Fitness :59.200000
Selected Gene Mean Fitness :59.400000
Selected Gene Mean Fitness :60.200000
Selected Gene Mean Fitness :61.200000
Selected Gene Mean Fitness :61.600000
Selected Gene Mean Fitness :62.400000
Selected Gene Mean Fitness :63.200000
Selected Gene Mean Fitness :63.600000
Selected Gene Mean Fitness :64.000000
Selected Gene Mean Fitness :64.200000
Selected Gene Mean Fitness :64.400000
Selected Gene Mean Fitness :64.800000
Selected Gene Mean Fitness :65.200000
Selected Gene Mean Fitness :65.600000
Selected Gene Mean Fitness :66.000000
Selected Gene Mean Fitness :66.200000
Selected Gene Mean Fitness :66.400000
Selected Gene Mean Fitness :66.600000
Selected Gene Mean Fitness :67.000000

result_9 = evolution_v2(geneSet, alpha=0.9)  # alpha 交叉系数

Selected Gene Group
Selected Gene Mean Fitness :52.200000
Selected Gene Mean Fitness :53.400000
Selected Gene Mean Fitness :54.400000
Selected Gene Mean Fitness :54.800000
Selected Gene Mean Fitness :56.200000
Selected Gene Mean Fitness :57.000000
Selected Gene Mean Fitness :57.400000
Selected Gene Mean Fitness :58.200000
Selected Gene Mean Fitness :59.000000
Selected Gene Mean Fitness :59.800000
Selected Gene Mean Fitness :60.000000
Selected Gene Mean Fitness :61.000000
Selected Gene Mean Fitness :62.400000
Selected Gene Mean Fitness :62.800000
Selected Gene Mean Fitness :63.800000
Selected Gene Mean Fitness :64.200000
Selected Gene Mean Fitness :64.400000
Selected Gene Mean Fitness :64.600000
Selected Gene Mean Fitness :64.800000
Selected Gene Mean Fitness :65.000000
Selected Gene Mean Fitness :65.200000
Selected Gene Mean Fitness :65.400000
Selected Gene Mean Fitness :65.600000
Selected Gene Mean Fitness :65.800000
Selected Gene Mean Fitness :66.400000
Selected Gene Mean Fitness :66.800000
Selected Gene Mean Fitness :67.000000

进化过程
在这里插入图片描述

从上图可以看出,进化过程陷入了局部最优,尽管尝试的不同的交叉比例,仍是没有改善,下面是经过选择后的TopN Gene图谱
在这里插入图片描述

这个图也说明,在单纯的交叉变异下,经过几轮进化后,父辈的基因基本趋于一致了,所以继续交叉下去是徒劳,很难到达最优.为了克服这个问题,继续尝试策略3,在交叉后进行突变.

3.交叉+突变

def cross_mutate(GeneGroup,alpha=0.3):
    childGenes = []
    for couple in itertools.combinations(GeneGroup, 2):
        childGene = couple[0].copy()
        length = len(childGene)
        random_location = random.sample(list(np.arange(length)), int(length*alpha))
        childGene[random_location] = couple[1][random_location] # 交叉
        childGene = unisexual_mutate(childGene, [0, 1])
        childGenes.append(childGene)
    return childGenes

def evolution_v3(geneSet, target=100, group_size=10, topN=5, alpha=0.5):
    gene_list = []
    mean_fitness = []
    max_fitness = []
    ParentGeneGroup = generate_parents_group(target, geneSet,group_size)
    #ParentGeneGroup = bisexual_mutate(ParentGeneGroup,alpha=0.2)
    selectedPGroup = select_topN_gene(ParentGeneGroup, topN)
    for gene in selectedPGroup:
        display(gene)
    print('Selected Gene Group')
    first_fitness = [get_fitness(gene) for gene in selectedPGroup]
    gene_list.append(selectedPGroup)
    mean_fitness.append(np.mean(first_fitness))
    max_fitness.append(np.max(first_fitness))
    # 
    for fitness in first_fitness:
        if fitness >= target:
            return selected_group
    #while True:
    for i in range(500):
        # 生成下一代
        childGeneGroup = cross_mutate(selectedPGroup,alpha=alpha)
        # 筛选
        selectedCGroup = select_topN_gene(childGeneGroup)
        gene_list.append(selectedCGroup) # collect
        # 上一代和下一代平均适应度进行比较
        mean_c_fitness = np.mean([get_fitness(gene) for gene in selectedCGroup])
        mean_p_fitness = np.mean([get_fitness(gene) for gene in selectedPGroup])
        max_c_fitness = np.max([get_fitness(gene) for gene in selectedCGroup])
        mean_fitness.append(mean_c_fitness)
        max_fitness.append(max_c_fitness)
        if mean_c_fitness <= mean_p_fitness:
            continue
        print('Selected Gene Mean Fitness :%f'%mean_c_fitness)
        if max_c_fitness >= target:
            return selectedCGroup, gene_list, mean_fitness, max_fitness
        selectedPGroup = selectedCGroup
    return selectedCGroup, gene_list, mean_fitness, max_fitness

bestGeneGroup, gene_list, mean_fitness, max_fitness = evolution_v3(geneSet=[0, 1], alpha=0.3)

gene: 1011111010100111000010110011011110100100010101011011110110111001101001101111101101000101010110111101, fitness: 59
gene: 0111111001011001101110101101101100011000111011011011010000000011101010111101110000001111111111011001, fitness: 58
gene: 0110010010000101110000110101011000001111110011101111110010011101111110101111100110010010010101000110, fitness: 54
gene: 0110010100000110110111011001111010011100111010110101001110000110111110111101101010011101000000010001, fitness: 53
gene: 0101011111000011000101011111010011111110101001111010000001000000001001101111111010000100111110101011, fitness: 53
Selected Gene Group
Selected Gene Mean Fitness :58.800000
Selected Gene Mean Fitness :62.200000
Selected Gene Mean Fitness :67.200000
Selected Gene Mean Fitness :69.600000
Selected Gene Mean Fitness :70.400000
Selected Gene Mean Fitness :73.200000
Selected Gene Mean Fitness :74.400000
Selected Gene Mean Fitness :76.000000
Selected Gene Mean Fitness :78.200000
Selected Gene Mean Fitness :80.600000
Selected Gene Mean Fitness :81.800000
Selected Gene Mean Fitness :82.600000
Selected Gene Mean Fitness :82.800000
Selected Gene Mean Fitness :84.000000
Selected Gene Mean Fitness :84.200000
Selected Gene Mean Fitness :85.000000
Selected Gene Mean Fitness :85.800000
Selected Gene Mean Fitness :87.200000
Selected Gene Mean Fitness :87.800000
Selected Gene Mean Fitness :88.200000
Selected Gene Mean Fitness :88.800000
Selected Gene Mean Fitness :90.200000
Selected Gene Mean Fitness :90.800000
Selected Gene Mean Fitness :91.000000
Selected Gene Mean Fitness :91.800000
Selected Gene Mean Fitness :92.200000
Selected Gene Mean Fitness :92.400000
Selected Gene Mean Fitness :92.600000
Selected Gene Mean Fitness :93.400000
Selected Gene Mean Fitness :93.800000
Selected Gene Mean Fitness :94.200000
Selected Gene Mean Fitness :94.600000
Selected Gene Mean Fitness :94.800000
Selected Gene Mean Fitness :95.200000
Selected Gene Mean Fitness :95.600000
Selected Gene Mean Fitness :96.000000
Selected Gene Mean Fitness :96.200000
Selected Gene Mean Fitness :96.400000
Selected Gene Mean Fitness :97.200000
Selected Gene Mean Fitness :97.400000
Selected Gene Mean Fitness :97.600000
Selected Gene Mean Fitness :97.800000
Selected Gene Mean Fitness :98.200000
Selected Gene Mean Fitness :98.400000
Selected Gene Mean Fitness :98.800000

进化过程
在这里插入图片描述

很明显,增加随机突变后,只需要更少的进化就达到了target.

对比

在这里插入图片描述

在这里插入图片描述

单个Gene的随机突变虽然也能到达目标,但是需要很长时间的进化,在之前的"Hello World"中,也是基于个体突变,但是这个不是完整遗传算法,算是精简版,策略3也不是完全的遗传算法,在选择过程中没有基于概率选择,而是采取了贪心的策略,每次都选择绝对的topN,交叉过程中也没有严格遵循遗传算法,而是选择了随机交叉.不过这些细节都不重要.重要的是,单纯的交叉可能会陷入局部最优,单纯突变可能要很多代的进化才行,交叉+突变可以缩短进化的时间.

计算机视觉领域优化算法综述：最优化、遗传算法、模拟退火算法、粒子群算法等
AI天才研究院
08-07 1154
在计算机视觉领域，优化算法一直是研究热点。很多研究人员都对优化算法进行了广泛研究，涉及最优化、遗传算法、模拟退火算法、粒子群算法等等。然而，对于如何选取适合CV任务的优化算法却鲜有论文进行详细分析。因此，本文提出了一系列理论和实践指导，帮助读者理解并选择适合于Computer Vision(CV)任务的优化算法。作者首先简要介绍了优化问题的类型、目标函数和约束条件，之后通过基于这些特征将优化算法分成几个重要的子类别，例如，局部搜索、迭代法、全局搜索以及优化模型。
使用 DEAP 库实现遗传算法解决旅行商问题
盼小辉丶的博客
08-28 2081
使用遗传算法 (Genetic Algorithm, GA) 解决旅行商问题 (Traveling Salesman Problem, TSP) 是一种常见且有效的方法。遗传算法模拟了生物进化的过程，通过交叉、变异和选择等操作来逐步优化问题的解。遗传算法作为一种启发式算法，虽然不能保证找到全局最优解，但通常能在合理时间内找到很好的近似解，尤其适合解决大规模的 TSP 问题。
DEAP Example: One Max Problem
年长的小白
08-09 1918
DEAP Sample: One Max Problem DEAP原文链接 本文是一个十分简单的入门任务。这个任务是求一个元素为0或1构成的数组的和的最大值。显然当元素全部为1的时候，数组的和有最大值，即为其长度。 Setting Things Up 首先，导入一些模块 import random from deap import base from deap import c...
DEAP Example: One Max Problem: Short Version
年长的小白
08-20 585
DEAP Example: One Max Problem: Short Version [TOC] 这个简短的例子与之前完整的One Max例子很相似。唯一区别是它采用了实现基本进化算法的算法模块。初始化基本相同，不同的是我们需要导入一些额外的包和模块。 import array import numpy from deap import algorithms 为了实现算法中的进化...
遗传算法求解one_max问题
yiyehuajia1993的专栏
01-29 2247
遗传算法求解one_max问题 one_max问题： 所谓的one_max就是指定任意长度的二进制串，当二进制串中的所有二进制数字都为1时，one_max就得解。用数学方法表示如下：（假设二进制串的长度为L） 也即最大化函数f(x)。 遗传算法： 遗传算法是模拟自然界生物的交配，基因的重组以及变异还有自然选择来进行求解最优化问题的一种算法。其间有大体包括genetic a
分布估计算法解决100维的OneMax问题
05-19
分布估计算法解决100维的OneMax问题，java写的，eclipse直接能跑
利用遗传算法找极小值_工程优化设计与Matlab实现——基本遗传算法求解一维问题...
weixin_31190007的博客
01-12 714
从猴子到人，本来就是一个优化过程。遗传算法Part 1 遗传算法的自然界原型遗传算法根据进化论的思想，模拟了生物界遗传选择和自然淘汰的生物进化过程。Point 1 从生物遗传谈起先回想一下生物遗传知识：基因是遗传物质最基本的结构和功能单位基因在染色体上基因是有遗传效应的DNA片段基因是如何在生物之间遗传的：假设某生物种群有n个个体；其中第n个个体有m个染色体Cn1、Cn2，……，Cnm；第n个个体...
一维下料问题的遗传算法基础：深度解析与应用指南
![一维下料问题的遗传算法基础：深度解析与应用指南]...本文首先概述了遗传算法及其在解决一维下料问题中的应用，随后深入探讨了遗传算法的基本理论和工作流程，
遗传算法(Python) #3 从零开始解决OneMax问题
youyanyu3817的博客
05-20 1281
遗传算法(Python) #3 从零开始解决OneMax问题 1. OneMax问题(OneMax Problem) OneMax问题是遗传算法的入门问题，其内容是：如何使一段长度固定的二进制字符串所有位置上数字之和最大。 让我们用一个长度为5的二进制字符串为例： 10010 -> 和为2 00111 -> 和为3 11111 -> 和为5(最大值) 对一般人，显而易见，当所有位数都为1时，该字符串的和最大，但在我们用遗传算法解决该问题时，遗传算法本身并没有这样的知识。接下来我们将不依
OneMax:具有一维最大问题的强化学习的遗传算法
05-11
OneMax 针对One Max问题进行强化学习的遗传算法包括实时图形，以查看GA的结果和参数+用法语撰写的有关该算法的文章以及开发“ presentation-des-algorithmes.pdf” + .jar进行启动的研究很容易
OneMaxOptimal:第三年实习源代码，用于计算最佳参数以解决oneMax问题-one source code
03-24
OneMax最佳 源代码，用于计算不同算法的最佳（和其他类型）参数，以解决OneMax问题 宾瓦尔 您将在此目录中找到一些代码来计算BinVal问题的最佳值。 实验结果 您将在此处找到用于根据Summary中计算出的参数计算经验数据的代码。 最佳EA 这是（1 + 1）EA的100到10000个最佳参数的所有结果。 最佳RLS 这是RLS最佳参数从100到10000的所有结果。 概括 有趣的是，这里存储了所有代码（以及结果），以针对不同算法计算最优参数或漂移最大化参数。
python 遗传算法_遗传算法(Python) #5 用DEAP框架解决OneMax问题
weixin_39846364的博客
12-05 354
遗传算法系列的第三期介绍了如何不用任何框架从零开始解决OneMax问题，第四期介绍了DEAP框架的基本用法。若读者对下文中定义或术语不熟悉，可以查看本系列的前几篇文章。本期文章将介绍如何使用DEAP解决OneMax问题。1.OneMax问题(OneMax problem)OneMax问题是遗传算法的入门问题，其内容是：如何使一段长度固定的二进制字符串所有位置上数字之和最大。让我们用一个长度为5的二...
使用 DEAP 库实现遗传算法解决 OneMax 问题
盼小辉丶的博客
08-19 1973
OneMax 问题是遗传算法和进化计算领域中的一个简单但经典的问题，常用于展示如何应用进化算法进行优化任务。在本节中，我们介绍了 DEAP 框架下遗传算法构建的通用流程，并使用 DEAP 解决简单的 OneMax 问题。
c遗传算法的终止条件一般_遗传算法(Python) #3 从零开始解决OneMax问题
weixin_36143786的博客
12-11 461
上一期1. OneMax问题(OneMax Problem)OneMax问题是遗传算法的入门问题，其内容是：如何使一段长度固定的二进制字符串所有位置上数字之和最大。让我们用一个长度为5的二进制字符串为例：10010 -> 和为200111 -> 和为311111 -> 和为5(最大值)对一般人，显而易见，当所有位数都为1时，该字符串的和最大，但在我们用遗传算法解决该问题时，遗传算...
Python中基于DEAP框架的遗传算法实现与OneMax问题求解
最新发布
参考资料请自行甄别 资源下载请先试读页面内容 观看内容概要 确认具体需求后再下载 亦或联系博主本人 不提供代码调试服务 如有疑问不解之处 请及时联系博主本人 妥善解决 加油 谢谢 你的鼓励是我前行的动力 谢谢支持 加油 谢谢
11-10 1072
Python中基于DEAP框架的遗传算法实现与OneMax问题求解（包含详细的完整的程序和数据）_遗传算法deap库资源-CSDN文库 https://download.csdn.net/download/xiaoxingkongyuxi/89872719。：初始化种群，评估适应度，进行选择、交叉和变异，不断进化，输出当前最佳适应度。：设置遗传算法的参数，如循环次数、种群大小、变异概率和交叉概率。对于较复杂的问题，可能需要调整遗传算法的其他要素。框架实现遗传算法，并包含详细的代码示例和解释。
遗传算法解决vrp问题_遗传算法(Python) #3 从零开始解决OneMax问题
weixin_31175771的博客
01-28 1681
上一期1. OneMax问题(OneMax Problem)OneMax问题是遗传算法的入门问题，其内容是：如何使一段长度固定的二进制字符串所有位置上数字之和最大。让我们用一个长度为5的二进制字符串为例：10010 -> 和为200111 -> 和为311111 -> 和为5(最大值)对一般人，显而易见，当所有位数都为1时，该字符串的和最大，但在我们用遗传算法解决该问题时，遗传算...
Python中从零开始的简单遗传算法
Together_CZ的博客
03-04 676
【翻译自 ：Differential Evolution Global Optimization With Python】 【说明：Jason BrownleePhD大神的文章个人很喜欢，所以闲暇时间里会做一点翻译和学习实践的工作，这里是相应工作的实践记录，希望能帮到有需要的人！】 遗传算法是一种随机全局优化算法。连同人工神经网络，它可能是最流行和广为人知的生物学启发算法之一。该算法是一种进化算法，它通过自然选择，具有二进制表示形式和基于遗传重组和遗传突...
遗传算法 (Genetic Algorithm, GA)
热门推荐
weixin_56577499的博客
04-24 2万+
遗传算法（Genetic Algorithm，简称GA）起源于对生物系统所进行的计算机模拟研究，是一种随机全局搜索优化方法，它模拟了自然选择和遗传中发生的复制、交叉(crossover)和变异(mutation)等现象，从任一初始种群（Population）出发，通过随机选择、交叉和变异操作，产生一群更适合环境的个体，使群体进化到搜索空间中越来越好的区域，这样一代一代不断繁衍进化，最后收敛到一群最适应环境的个体（Individual），从而求得问题的优质解。
遗传算法pyhton
10-16
遗传算法（Genetic Algorithm, GA）是一种模拟自然选择过程的优化搜索算法[^1]。在Python中，我们可以用它来解决旅行商问题（Traveling Salesman Problem, TSP）。以下是一个基本的遗传算法解决TSP问题的示例： 1. **初始化**：创建一个初始种群，每个个体代表一个可能的解决方案（路线），通常通过随机生成。 ```python import random def generate_route(n_cities): route = [i for i in range(1, n_cities + 1)] random.shuffle(route) return route initial_population = [generate_route(n_cities) for _ in range(population_size)] ``` 2. **适应度评估**：计算每个个体（路线）的总距离，作为其适应度。 ```python def fitness(route, distance_matrix): total_distance = sum(distance_matrix[i][route[i % len(route)]] for i in range(len(route))) return 1 / total_distance fitness_values = [fitness(route, distance_matrix) for route in initial_population] ``` 3. **选择**：基于适应度选择部分个体进入下一代。常用的选择策略有轮盘赌选择（roulette wheel selection）或锦标赛选择（tournament selection）。 4. **交叉**：通过交叉操作结合父母的基因，产生新的个体。常用的交叉方式有单点交叉（one-point crossover）。 ```python def single_point_crossover(parent1, parent2, crossover_point): child1 = parent1[:crossover_point] + parent2[crossover_point:] child2 = parent2[:crossover_point] + parent1[crossover_point:] return child1, child2 ``` 5. **变异**：对新产生的个体进行变异，增加多样性。如交换两个位置的操作（uniform crossover 或 bit-flip mutation）。 6. **重复**：直到达到停止条件（如迭代次数或适应度阈值），或者新一代的最优解优于上一代。 ```python while not stopping_condition: next_generation = [] ... (selection, crossover, and mutation steps are performed here ...) # 检查是否有新的最优解 if any(fitness(new_gen, distance_matrix) > best_fitness for new_gen in next_generation): best_solution = max(next_generation, key=fitness) ``` 这个框架展示了遗传算法的基本概念和实施步骤，但实际应用时需要细化并调整参数以获得更好的性能。
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值