2.3 均差与牛顿插值多项式
1.均差定义:定义
f
[
x
0
,
x
k
]
=
f
(
x
k
)
−
f
(
x
0
)
x
k
−
x
0
f[x_0,x_k]=\frac{f(x_k)-f(x_0)}{x_k-x_0}
f[x0,xk]=xk−x0f(xk)−f(x0)为函数
f
(
x
)
f(x)
f(x)关于点
x
0
x_0
x0,
x
k
x_k
xk的一阶均差,定义
f
[
x
0
,
x
1
,
x
k
]
=
f
[
x
0
,
x
k
]
−
f
[
x
0
,
x
1
]
x
k
−
x
1
f[x_0,x_1,x_k]=\frac{f[x_0,x_k]-f[x_0,x_1]}{x_k-x_1}
f[x0,x1,xk]=xk−x1f[x0,xk]−f[x0,x1]为函数
f
(
x
)
f(x)
f(x)关于点
x
0
x_0
x0,
x
1
x_1
x1,
x
k
x_k
xk的二阶均差,则定义
f
[
x
0
,
x
1
,
⋯
,
x
k
]
=
f
[
x
0
,
x
1
,
⋯
,
x
k
−
2
,
x
k
]
−
f
[
x
0
,
x
1
,
⋯
,
x
k
−
2
,
x
k
−
1
]
x
k
−
x
k
−
1
=
∑
j
=
0
n
f
(
x
j
)
ω
n
′
(
x
j
)
(1.17)
f[x_0,x_1,\cdots,x_k]=\frac{f[x_0,x_1,\cdots,x_{k-2},x_k]-f[x_0,x_1,\cdots,x_{k-2},x_{k-1}]}{x_k-x_{k-1}}=\sum_{j=0}^{n}\frac{f(x_j)}{\omega^\prime_n(x_j)} \tag{1.17}
f[x0,x1,⋯,xk]=xk−xk−1f[x0,x1,⋯,xk−2,xk]−f[x0,x1,⋯,xk−2,xk−1]=j=0∑nωn′(xj)f(xj)(1.17)
为函数
f
(
x
)
f(x)
f(x)关于
x
0
,
x
1
,
,
⋯
,
x
k
x_0,x_1,,\cdots,x_k
x0,x1,,⋯,xk的k阶均差(k阶差商)
2.均差性质:
(1)若
f
(
x
)
f(x)
f(x)在
[
a
,
b
]
[a,b]
[a,b]上存在
n
n
n阶导数,且节点
x
0
,
x
1
,
⋯
,
x
n
∈
[
a
,
b
]
x_0,x_1,\cdots,x_n\in[a,b]
x0,x1,⋯,xn∈[a,b],则n阶均差与导数关系
f
[
x
0
,
x
1
,
⋯
,
x
n
]
=
f
(
n
)
(
ξ
)
n
!
,
ξ
∈
[
a
,
b
]
(1.18)
f[x_0,x_1,\cdots,x_n]=\frac{f^{(n)}(\xi)}{n!},\xi \in[a,b]\tag{1.18}
f[x0,x1,⋯,xn]=n!f(n)(ξ),ξ∈[a,b](1.18)
(可采用罗尔定理证明)
(2)均差对称性
f
[
x
0
,
x
1
,
⋯
,
x
k
]
=
f
[
x
1
,
x
0
,
x
2
,
⋯
,
x
k
]
=
⋯
=
f
[
x
1
,
⋯
,
x
k
,
x
0
]
f[x_0,x_1,\cdots,x_k]=f[x_1,x_0,x_2,\cdots,x_k]=\cdots=f[x_1,\cdots,x_k,x_0]
f[x0,x1,⋯,xk]=f[x1,x0,x2,⋯,xk]=⋯=f[x1,⋯,xk,x0]
(3)由(1.17)式子和均差性质(2)得
f
[
x
0
,
x
1
,
⋯
,
x
k
]
=
f
[
x
0
,
x
1
,
⋯
,
x
k
−
2
,
x
k
]
−
f
[
x
0
,
x
1
,
⋯
,
x
k
−
2
,
x
k
−
1
]
x
k
−
x
k
−
1
(1.17-C)
f[x_0,x_1,\cdots,x_k]=\frac{f[x_0,x_1,\cdots,x_{k-2},x_k]-f[x_0,x_1,\cdots,x_{k-2},x_{k-1}]}{x_k-x_{k-1}}\tag{1.17-C}
f[x0,x1,⋯,xk]=xk−xk−1f[x0,x1,⋯,xk−2,xk]−f[x0,x1,⋯,xk−2,xk−1](1.17-C)
3.均差形式的牛顿插值多项式(插值节点任意分布情况使用)
由一次插值多项式得
P
1
(
x
)
=
P
0
(
x
)
+
f
[
x
0
,
x
1
]
(
x
−
x
0
)
=
f
(
x
0
)
+
f
[
x
0
,
x
1
]
(
x
−
x
0
)
\begin{aligned} P_1(x)&=P_0(x)+f[x_0,x_1](x-x_0)\\&=f(x_0)+f[x_0,x_1](x-x_0) \end{aligned}
P1(x)=P0(x)+f[x0,x1](x−x0)=f(x0)+f[x0,x1](x−x0)
由二次插值多项式得
P
2
(
x
)
=
P
1
(
x
)
+
f
[
x
0
,
x
1
,
x
2
]
(
x
−
x
0
)
(
x
−
x
1
)
=
f
(
x
0
)
+
f
[
x
0
,
x
1
]
(
x
−
x
0
)
(
x
−
x
1
)
\begin{aligned} P_2(x)&=P_1(x)+f[x_0,x_1,x_2](x-x_0)(x-x_1)\\ &=f(x_0)+f[x_0,x_1](x-x_0)(x-x_1) \end{aligned}
P2(x)=P1(x)+f[x0,x1,x2](x−x0)(x−x1)=f(x0)+f[x0,x1](x−x0)(x−x1)
根据均差定义,将
x
x
x看成
[
a
,
b
]
[a,b]
[a,b]上的一点,得
f
(
x
)
=
f
(
x
0
)
+
f
[
x
,
x
0
]
(
x
−
x
0
)
f
[
x
,
x
0
]
=
f
[
x
0
,
x
1
]
+
f
[
x
,
x
0
,
x
1
]
(
x
−
x
1
)
⋮
f
[
x
,
x
0
,
⋯
,
x
n
−
1
]
=
f
[
x
0
,
x
1
,
⋯
,
x
n
]
+
f
[
x
,
x
0
,
⋯
,
x
n
]
(
x
−
x
n
)
f(x)=f(x_0)+f[x,x_0](x-x_0)\\ f[x,x_0]=f[x_0,x_1]+f[x,x_0,x_1](x-x_1)\\ \vdots\\f[x,x_0,\cdots,x_{n-1}]=f[x_0,x_1,\cdots,x_n]+f[x,x_0,\cdots,x_n](x-x_n)
f(x)=f(x0)+f[x,x0](x−x0)f[x,x0]=f[x0,x1]+f[x,x0,x1](x−x1)⋮f[x,x0,⋯,xn−1]=f[x0,x1,⋯,xn]+f[x,x0,⋯,xn](x−xn)
叠加后得
f
(
x
)
=
f
(
x
0
)
+
f
[
x
0
,
x
1
]
(
x
−
x
0
)
+
f
[
x
0
,
x
1
,
x
2
]
(
x
−
x
0
)
(
x
−
x
1
)
+
⋯
+
f
[
x
0
,
x
1
,
⋯
,
x
n
]
(
x
−
x
0
)
⋯
(
x
−
x
n
−
1
)
+
f
[
x
,
x
0
,
⋯
,
x
n
]
ω
n
+
1
(
x
)
=
P
n
(
x
)
+
R
n
(
x
)
\begin{aligned} f(x)&=f(x_0)+f[x_0,x_1](x-x_0)+f[x_0,x_1,x_2](x-x_0)(x-x_1)+\cdots\\ &+f[x_0,x_1,\cdots,x_n](x-x_0)\cdots(x-x_{n-1})+f[x,x_0,\cdots,x_n]\omega_{n+1}(x)\\&=P_n(x)+R_n(x) \end{aligned}
f(x)=f(x0)+f[x0,x1](x−x0)+f[x0,x1,x2](x−x0)(x−x1)+⋯+f[x0,x1,⋯,xn](x−x0)⋯(x−xn−1)+f[x,x0,⋯,xn]ωn+1(x)=Pn(x)+Rn(x)
其中
P
n
(
x
)
=
f
(
x
0
)
+
f
[
x
0
,
x
1
]
(
x
−
x
0
)
+
f
[
x
0
,
x
1
,
x
2
]
(
x
−
x
0
)
(
x
−
x
1
)
+
⋯
+
f
[
x
0
,
x
1
,
⋯
,
x
n
]
(
x
−
x
0
)
⋯
(
x
−
x
n
−
1
)
(1.19)
\begin{aligned} P_n(x)&=f(x_0)+f[x_0,x_1](x-x_0)+f[x_0,x_1,x_2](x-x_0)(x-x_1)+\cdots\\ &+f[x_0,x_1,\cdots,x_n](x-x_0)\cdots(x-x_{n-1}) \end{aligned}\tag{1.19}
Pn(x)=f(x0)+f[x0,x1](x−x0)+f[x0,x1,x2](x−x0)(x−x1)+⋯+f[x0,x1,⋯,xn](x−x0)⋯(x−xn−1)(1.19)
R n ( x ) = f ( x ) − P n ( x ) = f [ x , x 0 , ⋯ , x n ] ω n + 1 ( x ) (1.20) \begin{aligned} R_n(x)=f(x)-P_n(x)=f[x,x_0,\cdots,x_n]\omega_{n+1}(x)\end{aligned}\tag{1.20} Rn(x)=f(x)−Pn(x)=f[x,x0,⋯,xn]ωn+1(x)(1.20)
对于多项式
P
n
(
x
)
P_n(x)
Pn(x)满足
P
n
(
x
i
)
=
f
(
x
i
)
P_n(x_i)=f(x_i)
Pn(xi)=f(xi),且次数不超过n,将
P
n
(
x
)
P_n(x)
Pn(x)表示为
P
n
(
x
)
=
a
0
+
a
1
(
x
−
x
0
)
+
⋯
+
a
n
(
x
−
x
0
)
⋯
(
x
−
x
n
−
1
)
P_n(x)=a_0+a_1(x-x_0)+\cdots+a_n(x-x_0)\cdots(x-x_{n-1})
Pn(x)=a0+a1(x−x0)+⋯+an(x−x0)⋯(x−xn−1)
其中
a
k
=
f
[
x
0
,
x
1
,
⋯
,
x
n
]
,
k
=
0
,
1
,
⋯
,
n
a_k=f[x_0,x_1,\cdots,x_n],k=0,1,\cdots,n
ak=f[x0,x1,⋯,xn],k=0,1,⋯,n
称
P
n
(
x
)
P_n(x)
Pn(x)为牛顿均差插值多项式,余项为
R
n
(
x
)
=
f
(
x
)
−
P
n
(
x
)
=
f
[
x
0
,
x
1
,
⋯
,
x
n
]
ω
n
+
1
(
x
)
(1.21)
R_n(x)=f(x)-P_n(x)=f[x_0,x_1,\cdots,x_n]\omega_{n+1}(x)\tag{1.21}
Rn(x)=f(x)−Pn(x)=f[x0,x1,⋯,xn]ωn+1(x)(1.21)
4.差分形式的牛顿均值多项式(等距节点分布情况使用)
对于
x
k
=
x
0
+
k
h
x_k=x_0+kh
xk=x0+kh此类的节点分布,
h
h
h为步长;设
x
k
x_k
xk对应的函数值
f
k
=
f
(
x
k
)
,
k
=
0
,
1
,
⋯
,
n
f_k=f(x_k),k=0,1,\cdots,n
fk=f(xk),k=0,1,⋯,n,称
Δ
f
k
=
f
k
+
1
−
f
k
\Delta f_k=f_{k+1}-f_k
Δfk=fk+1−fk为
x
k
x_k
xk处以
h
h
h为步长的一阶(向前)差分;称
Δ
2
f
k
=
Δ
f
k
+
1
−
Δ
f
k
\Delta^2 f_k=\Delta f_{k+1}-\Delta f_k
Δ2fk=Δfk+1−Δfk为
x
k
x_k
xk处的二阶差分,一般地称
Δ
n
f
k
=
Δ
n
−
1
f
k
+
1
−
Δ
n
−
1
f
k
(1.22)
\Delta^n f_k=\Delta^{n-1}f_{k+1}-\Delta^{n-1} f_k\tag{1.22}
Δnfk=Δn−1fk+1−Δn−1fk(1.22)
为
x
k
x_k
xk处的n阶差分
定义
I
f
k
=
f
k
,
E
f
k
=
f
k
+
1
If_k=f_k,Ef_k=f_{k+1}
Ifk=fk,Efk=fk+1,其中
I
I
I为不变算子,
E
E
E为步长为
h
h
h的位移算子
Δ
f
k
=
f
k
+
1
−
f
k
=
E
f
k
−
I
f
k
=
(
E
−
I
)
f
k
\Delta f_k=f_{k+1}-f_k=Ef_k-If_k=(E-I)f_k
Δfk=fk+1−fk=Efk−Ifk=(E−I)fk
Δ n f k = ∑ j = 0 n ( − 1 ) j ( n j ) E n − j f k = ∑ j = 0 n ( − 1 ) j ( n j ) f n + k − j (1.23) \Delta^n f_k=\sum_{j=0}^{n}(-1)^j\begin{pmatrix}n\\j\end{pmatrix}E^{n-j}f_k =\sum_{j=0}^{n}(-1)^j\begin{pmatrix}n\\j\end{pmatrix}f_{n+k-j}\tag{1.23} Δnfk=j=0∑n(−1)j(nj)En−jfk=j=0∑n(−1)j(nj)fn+k−j(1.23)
其中
(
n
j
)
=
n
(
n
−
1
)
⋯
(
n
−
j
+
1
)
j
!
\begin{pmatrix} n\\j \end{pmatrix}=\frac{n(n-1)\cdots(n-j+1)}{j!}
(nj)=j!n(n−1)⋯(n−j+1)为二项式展开系数,由此(1.21)式能够将各阶差分通过函数值给出;而也可以用各阶差分表示函数值
f
n
+
k
=
E
n
f
k
=
(
I
+
Δ
)
n
f
k
=
∑
j
=
0
n
(
n
j
)
Δ
i
f
k
(1.24)
f_{n+k}=E^nf_k=(I+\Delta)^n f_k=\sum_{j=0}^{n}\begin{pmatrix}n\\j\end{pmatrix}\Delta^if_k\tag{1.24}
fn+k=Enfk=(I+Δ)nfk=j=0∑n(nj)Δifk(1.24)
均差和差分的关系
f
[
x
k
,
⋯
,
x
k
+
m
]
=
1
m
!
1
h
m
Δ
m
f
k
,
m
=
1
,
2
,
⋯
,
n
(1.25)
f[x_k,\cdots,x_{k+m}]=\frac{1}{m!}\frac{1}{h^m}\Delta^mf_k,m=1,2,\cdots,n\tag{1.25}
f[xk,⋯,xk+m]=m!1hm1Δmfk,m=1,2,⋯,n(1.25)
由(1.18)式均差与导数的关系,得到差分与导数(微商)的关系为
Δ
n
f
k
=
h
n
f
n
(
ξ
)
,
ξ
∈
(
x
k
,
x
k
+
1
)
(1.26)
\Delta^nf_k=h^nf^{n}(\xi),\xi\in(x_k,x_{k+1})\tag{1.26}
Δnfk=hnfn(ξ),ξ∈(xk,xk+1)(1.26)
在(1.19)式中,均差部分都采用差分代替,即通过(1.25)式替换,且令
x
=
x
0
+
t
h
x=x_0+th
x=x0+th,得
P
n
(
x
)
=
P
n
(
x
0
+
t
h
)
=
f
0
+
t
Δ
f
0
+
t
(
t
−
1
)
2
!
Δ
2
f
0
+
⋯
+
t
(
t
−
1
)
⋯
(
t
−
n
+
1
)
n
!
Δ
n
f
0
(1.27)
P_n(x)=P_n(x_0+th)=f_0+t\Delta f_0+\frac{t(t-1)}{2!}\Delta^2f_0+\cdots\\+\frac{t(t-1)\cdots(t-n+1)}{n!}\Delta^nf_0\tag{1.27}
Pn(x)=Pn(x0+th)=f0+tΔf0+2!t(t−1)Δ2f0+⋯+n!t(t−1)⋯(t−n+1)Δnf0(1.27)
称(1.27)式为牛顿前插公式,余项为
R
n
(
x
)
=
f
(
x
)
−
P
n
(
x
)
=
t
(
t
−
1
)
⋯
(
t
−
n
)
(
n
+
1
)
!
h
n
+
1
f
n
+
1
(
ξ
)
,
ξ
∈
(
x
0
,
x
n
)
(1.28)
R_n(x)=f(x)-P_n(x)=\frac{t(t-1)\cdots(t-n)}{(n+1)!}h^{n+1}f^{n+1}(\xi),\xi\in(x_0,x_n)\tag{1.28}
Rn(x)=f(x)−Pn(x)=(n+1)!t(t−1)⋯(t−n)hn+1fn+1(ξ),ξ∈(x0,xn)(1.28)
误差分析时可以写成
∣
R
n
(
x
)
∣
≤
M
n
+
1
(
n
+
1
)
!
∣
t
(
t
−
1
)
⋯
(
t
−
n
)
∣
h
n
+
1
(1.29)
|R_n(x)|\leq \frac{M_{n+1}}{(n+1)!}|t(t-1)\cdots(t-n)|h^{n+1}\tag{1.29}
∣Rn(x)∣≤(n+1)!Mn+1∣t(t−1)⋯(t−n)∣hn+1(1.29)
其中
f
(
n
+
1
)
(
ξ
)
≤
M
n
+
1
f^{(n+1)}(\xi)\leq M_{n+1}
f(n+1)(ξ)≤Mn+1