遗传算法 学习笔记

（原文曾在：http://liyitan.sinaapp.com 都是本人）

最近听大牛讨论遗传算法，讨论其对各种启发搜索问题，都有比较好的鲁棒性，所以心痒+手贱，尝试了一下。

实际上仅仅只实现了，搜索一个字符串（有数字组成）的功能。Python代码如下（比较水）：

话所，速度果然还真够慢的······

输入：一个字符串（从data.txt文件读入），作为搜索目标

输出：每50次繁殖，显示一个种群中的最优解（就是距离）

注释：当然，这个问题本身自然不需要使用遗传算法，我只是写着玩

import matplotlib.pyplot as plt
import numpy as np
import copy

def similar(x,y):
    sum = 0
#    print x,y
    for i,elem in enumerate(x):
        sum += np.abs(elem-y[i])
    return sum

class gene:
    def __init__(self,data):
        self.data = data
    def sortkey(self,x):
        return similar(self.data,x)
    def select(self,Mlist,M):
        p = []
        for elem in Mlist:
            #the function which will effect the dztion
            p.append(float(1)/(1+similar(self.data,elem)))
        mu = np.sum(p)
        p = [elem/float(mu) for elem in p]
        newMlist = []
        se = -1
        while len(newMlist) < M:
            you = np.random.rand()
            for i,elem in enumerate(p):
                you -= elem
                if you < 0:
                    se = i
                    break
            newMlist.append(Mlist[se])
        return newMlist
    def min(a,b):
        if a<0 or b < 0:
            raise Exception('cccc')
        if a < b:
            return a
        return b
    def geneNum(self,M,pb,pj):
        Mlist = []
        for i in range(M):
            temp = []
            for j in range(len(self.data)):
                temp.append(int(np.random.rand()*10))
            Mlist.append(temp)
        count = 0
        while True:
            newMlist = []
            Mlist.sort(key = self.sortkey)
#            print 'sort'
            if count >= 1000 or similar(Mlist[0],self.data) == 0:
                print '%d:' % count +str(Mlist[0])
                break
            if count%50 == 0:
                print '%d:' % count +str(Mlist[0])
#            for i in range(M):
            newMlist = self.select(Mlist,M)
#                newMlist.append(Mlist[i])
            for i in range(int(M*pb)):
                temp = copy.copy(Mlist[int(np.random.rand()*len(Mlist))])
                chg = int(np.random.rand()*10)
                pos = int(np.random.rand()*len(self.data))
                temp[pos] = chg
                newMlist.append(temp)
#            print count
#            print 'b:'+str(newMlist)
            for i in range(int(M*pj)):
                a = int(np.random.rand()*len(Mlist))
                b = int(np.random.rand()*len(Mlist))
                pos1 = int(np.random.rand()*len(self.data))
                pos2 = int(np.random.rand()*len(self.data))
                length = int(np.random.rand()*min(len(self.data)-pos1-1,len(self.data)-pos2-1))
                temp = Mlist[a][:pos1]
                temp.extend(Mlist[b][pos2:pos2+length])
                temp.extend(Mlist[a][pos1+length:])
                newMlist.append(temp)
                temp = Mlist[b][:pos2]
                temp.extend(Mlist[a][pos1:pos1+length])
                temp.extend(Mlist[b][pos2+length:])
                newMlist.append(temp)
#            print 'j:'+str(newMlist)
            Mlist = newMlist
#            print newMlist
            count += 1
def main():
    with open('data.txt','rb') as f:
        sp = f.readline().split()
        data = []
        for c in sp[0]:
            data.append(int(c))
    g = gene(data)
    g.geneNum(100,0.1,0.7)

if __name__ == '__main__':
    main()