插值题目选做

n次多项式空间是有限维线性空间

◉ { x 0 , ⋯   , x n } \{x_0,\cdots,x_n\} { x0​,⋯,xn​}互异, k ∈ { 0 , ⋯   , n } k \in \{0,\cdots,n\} k∈{ 0,⋯,n}, 求证

1. ∑ i = 0 n x i k l i ( x ) ≡ x k , k ∈ { 0 , ⋯   , n } \sum\limits_{i=0}^{n} x_i^k l_i(x) \equiv x^k, \quad k \in \{0,\cdots,n\} i=0∑n​xik​li​(x)≡xk,k∈{ 0,⋯,n}
2. ∑ i = 0 n ( x i − x ) k l i ( x ) ≡ 0 , k ∈ { 1 , ⋯   , n } \sum\limits_{i=0}^{n} (x_i-x)^k l_i(x) \equiv 0, \quad k \in \{1,\cdots,n\} i=0∑n​(xi​−x)kli​(x)≡0,k∈{ 1,⋯,n}

1.

$x^k$是一个不超过$n$次的多项式.

大前提: 给定$x^k$上的$n+1$个相异的点进行插值, 由插值的唯一性可知插值多项式恒等于$x^k$.

小前提: 给定$x^k$上的$n+1$个相异的点( x i ( i ∈ { 0 , ⋯   , n } ) x_i(i \in \{0,\cdots,n\}) xi​(i∈{ 0,⋯,n}))进行(Lagrange)插值, $\sum _{i=0}^n{x}_i^k{l}_i(x)$是插值多项式.

结论: $\sum _{i=0}^n{x}_i^k{l}_i(x)$恒等于$x^k$.

2.

k ∈ { 0 , ⋯   , n } k \in \{0,\cdots,n\} k∈{ 0,⋯,n}, $\sum _{i=0}^n(x_i-t{)}^k{l}_i(x)\equiv (x-t{)}^k$.
令$t=x$, $\sum _{i=0}^n(x_i-x{)}^k{l}_i(x)\equiv (x-x{)}^k\equiv 0$.

插值余项

◉ 给出分段两点三次Hermite插值误差限的理论分析.

$2n+1$次插值余项

$a=x_0<\cdots <x_n=b$
$\hat{\pi }(x)=\prod _{i=0}^n(x-x_i{)}^2$

$R_{2n+1}(x)=f(x)-H_{2n+1}(x)$
$K(x)=\frac{R_{2n+1}(x)}{\hat{\pi }(x)}$
$\varphi (t)=R_{2n+1}(t)-K(x)\hat{\pi }(t)$

注意到$\varphi (t)$有$n+2$个零点 { x 0 , ⋯   , x n , x } \{x_0,\cdots,x_n,x\} { x0​,⋯,xn​,x}, ${\varphi }^{\prime }(t)$也有$n+1$个零点 { x 0 , ⋯   , x n } \{x_0,\cdots,x_n\} { x0​,⋯,xn​}, 所以$\varphi (t)$有$2n+3$个根($n+1$个二重根 { x 0 , ⋯   , x n } \{x_0,\cdots,x_n\} { x0​,⋯,xn​},$1$个一重根 { x } \{x\} { x}). 迭用Rolle定理得到${\varphi }^{(2n+2)}(\xi )=0$(其中$\xi \in (a,b)$).

$R_{2n+1}^{(2n+2)}(\xi )=f^{(2n+2)}(\xi )$
${\hat{\pi }}^{(2n+2)}(\xi )=(2n+2)!$
${\varphi }^{(2n+2)}(\xi )=f^{(2n+2)}(\xi )-K(x)(2n+2)!=0$
$K(x)=\frac{R_{2n+1}(x)}{\hat{\pi }(x)}=(2n+2$