【计算方法】实验四：python实现拉格朗日插值和牛顿插值

1. lagrange 插值拟合

import numpy as np
import matplotlib.pyplot as plt
from sympy import *
from sympy.plotting import plot
from sympy.abc import x
​
def lagrange(x,y,x0):
    n = len(x)
    Ln = 0
    for k in range(n):  #对插值点循环
        Tk = 1
        for j in range(n):
            if j!=k:
                Tk = Tk * (x0-x[j])/(x[k] - x[j])
        Ln = Ln + y[k]*Tk
    return Ln
​
x = np.arange(-5,6,1)
y = list(map(lambda i: 1/(1+i**2),x))
x0= symbols('x')
​
lagrange(x,y,x0)
simplify(lagrange(x,y,x0))

​
插值得到的多项式：

​

插值多项式与原函数

​

x_new = np.arange(-5, 6, 0.1)
f_result = []
x = symbols('x')
for i in x_new:
    f_result.append(f.subs(x,i)) # 把不同的x带入求值
#plt.scatter(x, y, label='data')
plt.plot(x_new, f_result, label='lagrange-nih',linestyle='-.')
plt.plot(x_new, 1/(1+x_new**2),label="Polynomial-100")
plt.legend()
plt.axis([-5,5,-2,2])
plt.show()

​

​

​
​

拉格朗日	插值多项式	原多项式
x = 3.5	−0.2261962890625	0.07547169811320754
x = 4.5	1.57872099034926	0.047058823529411764

2.newton插值

import numpy as np
import matplotlib.pyplot as plt
from sympy import *
from sympy.plotting import plot
from sympy.abc import x
​
def newton(x,y,x0):
    n=len(x)
    Nn = y[0]  #第一列存函数值
    fn = np.ones([n,n])  #初始化差商骨架结构
    fn[:,0] = y  #差商表第一列
    Nj = 1    #连乘的设置为1
​
    for k in range(1,n):  #从第二列开始求均差
        for j in range(k,n):
            fn[j,k] = (fn[j,k-1]-fn[j-1,k-1])/(x[j]-x[j-k]) #改x[j]-x[j-k+1]
        Nj = Nj*(x0-x[k-1])
        Nn = Nn+fn[k,k]*Nj
        
    return Nn
​
x = np.arange(-5,6,1)
y = list(map(lambda i: 1/(1+i**2),x))
x0= symbols('x')
​
f = newton(x,y,x0)
simplify(f)  

​
得到的多项式：

​

插值函数与原函数

​

x_new = np.arange(-5, 6, 0.1)
f_result = []
x = symbols('x')
for i in x_new:
    f_result.append(f.subs(x,i)) # 把不同的x带入求值
#plt.scatter(x, y, label='data')
plt.plot(x_new, f_result, label='newton-nih',linestyle='-.')
plt.plot(x_new, 1/(1+x_new**2),label="Polynomial-100")
plt.legend()
plt.axis([-5,5,-2,2])
plt.show()

​

牛顿	插值多项式	原多项式
x = 3.5	−0.226196289062507	0.07547169811320754
x = 4.5	1.57872099034924	0.047058823529411764

对于插值多项式得到的值发现与原多项式得到的值差别很大，即插值函数与数据点的趋势不同，这时由于拉格朗日插值的局限性而导出现了Range现象。出现了在等距点插值的高次多项式相关的极端的“多项式摆动”
​