罗孝麟_遗传算法.pptx

罗孝麟 2017.3.11;Genetic Algorithms;Step 1 individuals in the population are n-dimensional binary vectors x, and the goal is to minimize some objective function S(x).;Step 2 Determining the M “fittest” individuals could be via tournament selection. （1）In basic tournament selection with tournaments of size K, this involves selecting K individuals uniformly from the population （2）selecting the individual with the lowest objective function value as the winner. The winner then joins the reproduction pool. This process is repeated M times;Step 3 Combine with one-point crossover: given two parents xand y, and a random location r between 0 and n, create a new individual z = (x1, . . . , xr, yr+1, . . . , yn) ;Step 4 Mutation flipping each component ofeach binary vector independently with probability p = 1/n;Step 6 If a stopping criterion is met, stop There are so many conditions we can use A solution is found that satisfies minimum criteria Fixed number of generations reached Allocated budget (computation time/money) reached Manual inspection; The above-mentioned content comes from Monte Carlo Methods ,Dirk P.Kroese. There is a example about kangaroo on /article/details?id=6938608 ;问题可以简化为数值求解一个函数的全局最小值;“袋鼠跳”问题 我们把 函数曲线理解成一个一个山峰和山谷组成的山脉。那么我们可以设想所得到的每一个解就是一只袋鼠，我们希望它们不断的向着更高处跳去，直到跳到最高的山峰（尽管袋鼠本身不见得愿意那么做）。所以求最大值的过程就转化成一个“袋鼠跳”的过程。;以下是这个例子的求解过程;?;简介;1）变异方法;2）适应性函数;3）轮盘赌法;4）收敛条件; double x = fabs(p[0] ); double y = p[1]; double chi2 = ((x-1)*(x-1)-1)*((p[0]0)*2 +(p[0]=0)*0.5) + (y-2)*(y-2) + (p[2]-1)*(p[2]-1) + (p[3]-2)*(p[3]-2) + (p[4]-1)*(p[4]-1) + (p[5]-2)*(p[5]-2) + (p[6]-1)*(p[6]-1) + (p[7]-2)*(p[7]-2) ; return chi2;;小结