使用Python绘制一个非常漂亮的李雅普诺夫指数,及其函数值
import numpy as np import matplotlib.pyplot as plt # show plots in notebook # % matplotlib inline result = [] lambdas = [] maps = [] # define range of r rvalues = np.arange(0, 2, 0.01) # loop through r for r in rvalues: x = 0.1 result = [] # iterate system 100 times for t in range(100): x = x + r - x**2 # calculate log of the absolute of the derivative # 计算李雅普诺夫指数最关键的地方是,计算导数; result.append(np.log(abs(1 - 2*x))) # take average lambdas.append(np.mean(result)) # for the map ignore first 100 iterations as transient time and iterate anew for t in range(20): x = x + r - x**2 maps.append(x) fig = plt.figure(figsize=(10,7)) ax1 = fig.add_subplot(1,1,1) xticks = np.linspace(0, 2, 4000) # zero line zero = [0]*4000 ax1.plot(xticks, zero, 'g-') # plot map ax1.plot(xticks, maps, 'r.',alpha = 0.3, label = 'Map') ax1.set_xlabel('r') # plot lyapunov ax1.plot(rvalues, lambdas, 'b-', linewidth = 3, label = 'Lyapunov exponent') ax1.grid('on') ax1.set_xlabel('r') ax1.legend(loc='best') ax1.set_title('Map of x(t+1) = x(t) + r - x(t)^2 versus Lyapunov exponent') plt.show()
代码解析:
1、xticks的长度是4000,200*20;
2、一共for循环了200个不同的r值(即模型参数);
3、每个r值计算出来一个李雅普诺夫指数;
4、每个r值计算出来20个不同的幅值,(如果振动比较小,这个20个值,很接近),如果进入混沌态,这20个值差异就很大,这说明此时整个系统对模型参数的微小变化非常敏感;
5、