求遗传算法最高/平均适应度：Python中Map/Reduce的替代方案

需求说明：计算遗传算法中基因组的最高适应度与平均适应度，适应度函数为基因组字符串中1的数量，输入种群为：

['00111110011001010011', '01111101001101110010', '01100110111110000000', '01101101100111001001']

原实现代码 #

from functools import reduce

def fitness(genome):
    return reduce((lambda x, y: int(x) + int(y)), list(genome))

def evaluateFitness(population):
    sums = list(map(lambda x: fitness(x), population))
    return [max(sums), sum(sums) / len(population)]

# 测试
population = ['00111110011001010011', '01111101001101110010', '01100110111110000000', '01101101100111001001']
print(evaluateFitness(population))  # 输出: [12, 9.75]

1. Numpy实现方案 #

Numpy的向量化操作能高效处理这类字符串统计，无需手动写map/reduce：

import numpy as np

def evaluate_fitness_numpy(population):
    # 将所有基因组字符串转为二维整数数组
    genome_array = np.array([list(genome) for genome in population], dtype=int)
    # 每行求和得到单个基因组的适应度
    fitness_scores = genome_array.sum(axis=1)
    # 返回最高适应度和平均值
    return [fitness_scores.max(), fitness_scores.mean()]

# 测试
population = ['00111110011001010011', '01111101001101110010', '01100110111110000000', '01101101100111001001']
print(evaluate_fitness_numpy(population))  # 输出: [12, 9.75]

优势：种群规模大时，向量化运算比循环/映射更快，代码更简洁。

2. Scipy实现方案 #

Scipy没有专属的字符串统计函数，但可以结合Numpy+Scipy的统计工具完成任务，适合需要多维度统计信息的场景：

import numpy as np
from scipy.stats import describe

def evaluate_fitness_scipy(population):
    genome_array = np.array([list(genome) for genome in population], dtype=int)
    fitness_scores = genome_array.sum(axis=1)
    # 一次性获取最值、均值、方差等统计结果
    stats_result = describe(fitness_scores)
    return [stats_result.minmax[1], stats_result.mean]

# 测试
population = ['00111110011001010011', '01111101001101110010', '01100110111110000000', '01101101100111001001']
print(evaluate_fitness_scipy(population))  # 输出: [12, 9.75]

3. Pygad实现方案 #

Pygad是专门的遗传算法库，适合完整的遗传算法流程，内置适应度计算机制：

import pygad

def fitness_function(ga_instance, solution, solution_idx):
    # solution是Numpy数组，直接求和得到适应度
    return sum(solution)

# 把字符串基因组转为整数列表，适配Pygad的输入格式
population = [list(map(int, genome)) for genome in ['00111110011001010011', '01111101001101110010', '01100110111110000000', '01101101100111001001']]

# 初始化GA实例，仅用于评估当前种群（无需进化）
ga = pygad.GA(
    num_generations=1,
    num_parents_mating=2,
    fitness_func=fitness_function,
    sol_per_pop=len(population),
    initial_population=population,
    gene_type=int
)

# 计算种群适应度
ga.cal_pop_fitness()
fitness_scores = ga.pop_fitness

max_fitness = max(fitness_scores)
avg_fitness = sum(fitness_scores) / len(fitness_scores)
print([max_fitness, avg_fitness])  # 输出: [12, 9.75]

优势：如果后续需要进行选择、交叉、变异等遗传算法核心操作，Pygad能一站式完成，无需自己实现整套逻辑。

内容的提问来源于stack exchange，提问作者Shariq