今晚简单研究下遗传算法,学习的是一个求N个数,使之加起来恰好为X的例子,比较简明易懂,python实现起来也很方便。
几个基础的概念:
个体individual,它有自己的生存力,也就是适应力,强弱就是与我们的目标的差距
种群population,个体的集合,生存力不同,有强有弱
适应力fitness,针对个体而言,越小越好,看它与目标的差距
评分grade,针对种群而言,同样越小越好,定义了所有个体与目标差距的平均值
进化evolve,核心部分,生物的物竞天择适者生存过程,不断淘汰种群中的弱者,留下强者,并从强者中选择2个作为父母繁衍后代,后代有父母的基因,同时产生的过程中有概率发生变异,也可以选择让父母产生变异,从结果上看效果是一样的。
下面的例子中,设定目标值371,种群中个体数100,每个个体由6个数组成,在0到100之间,每次进化留下的优良个体比例20%,不良个体被留下的概率为5%(这个可以不要,留下会表现有遗传的多样性),留下的个体中,变异概率1%。进化前会对种群中个体的适应力排序,选择一定比例的留下,然后让其中的每个按概率发生变异,结果作为父母,繁衍后代,直到个体总量达到规定值。这里,我们预先知道我们的目标值,因此发现有个体完全适应时就可以停止进化了,而有些问题并不能准确知道这个值,因此可以将结果不断的保留,最后取一个最值作为我们的结果,得到原问题的近似最优解。
1 # -*- coding:gbk -*- 2 import random, operator 3 4 def individual(length, min, max): 5 return [random.randint(min, max) for x in xrange(length)] 6 7 def population(count, length, min, max): 8 return [individual(length, min, max) for x in xrange(count)] 9 10 def fitness(individual, target):11 sum = reduce(operator.add, individual, 0)12 return abs(target - sum)13 14 def grade(pop, target):15 summed = reduce(operator.add, (fitness(x, target) for x in pop))16 return summed / (len(pop) * 1.0)17 18 def evolve(pop, target, retain = 0.2, random_select = 0.05, mutate = 0.01):19 graded = [(fitness(x, target), x) for x in pop]20 graded = [x[1] for x in sorted(graded)]21 retain_length = int(len(graded) * retain)22 parents = graded[:retain_length]23 for individual in graded[retain_length:]:24 if random_select > random.random():25 parents.append(individual)26 27 for individual in parents:28 if mutate > random.random():29 pos_to_mutate = random.randint(0, len(individual) - 1)30 individual[pos_to_mutate] = random.randint(min(individual), max(individual))31 parents_length = len(parents)32 desired_length = len(pop) - parents_length33 children = []34 while len(children) < desired_length:35 male = random.randint(0, parents_length - 1)36 female = random.randint(0, parents_length - 1)37 if male != female:38 male = parents[male]39 female = parents[female]40 half = len(male) / 241 child = male[:half] + female[half:]42 children.append(child)43 parents.extend(children)44 return parents45 46 47 target = 37148 p_count = 10049 i_length = 650 i_min = 051 i_max = 10052 53 p = population(p_count, i_length, i_min, i_max)54 fitness_history = [grade(p, target),]55 for i in xrange(200):56 p = evolve(p, target)57 g = grade(p, target)58 fitness_history.append(g)59 if g == 0:60 break61 62 for datum in fitness_history:63 print datum64 65 individual = p[len(p) - 1]66 print 'individual is'67 sum = 068 for n in individual:69 sum += n70 print n71 print 'total=%d,target=%d,evolve=%d'%(len(fitness_history), target, sum)