注:本文的内容主要根据文末中的参考文档中的内容进行整理完成。
在微分方程的研究中,数学家提出Chebyshev微分方程:
( 1 − x 2 ) y ′ ′ − x y ′ + n 2 y = 0 , (1) (1-x^{2})y''-xy'+n^{2}y=0, \tag{1} (1−x2)y′′−xy′+n2y=0,(1)
( 1 − x 2 ) y ′ ′ − 3 x y ′ + n ( n + 2 ) y = 0 , (2) (1-x^{2})y''-3xy'+n(n+2)y=0, \tag{2} (1−x2)y′′−3xy′+n(n+2)y=0,(2)
第一类和第二类Chebyshev多项式分别对应上述两个方程的解。
两类Chebyshev多项式之间有如下关系:
d d x T n ( x ) = n U n − 1 ( x ) , n = 1 , ⋯ , T n ( x ) = 1 2 ( U n ( x ) − U n − 2 ( x ) ) , T n + 1 ( x ) = x T n ( x ) − ( 1 − x 2 ) U n − 1 ( x ) , T n ( x ) = U n ( x ) − x U n − 1 ( x ) . (9) \begin{aligned} & \frac{d}{dx}T_{n}(x) = nU_{n-1}(x), n=1,\cdots, \\ & T_{n}(x) = \frac{1}{2}(U_{n}(x)-U_{n-2}(x)),\\ & T_{n+1}(x) = xT_{n}(x)-(1-x^{2})U_{n-1}(x),\\ & T_{n}(x) = U_{n}(x)-xU_{n-1}(x). \tag{9} \end{aligned} dxdTn(x)=nUn−1(x),n=1,⋯,Tn(x)=21(Un(x)−Un−2(x)),Tn+1(x)=xTn(x)−(1−x2)Un−1(x),Tn(x)=Un(x)−xUn−1(x).(9)
对每个非负整数 n n n, T n ( x ) T_{n}(x) Tn(x)和 U n ( x ) U_{n}(x) Un(x)都为 n n n次多项式,并且
可用公式描述其奇偶性:
T n ( x ) = ( − 1 ) n T n ( − x ) , U n ( x ) = ( − 1 ) n U n ( − x ) . \begin{aligned} &T_{n}(x) = (-1)^{n}T_{n}(-x),\\ & U_{n}(x) = (-1)^{n}U_{n}(-x). \end{aligned} Tn(x)=(−1)nTn(−x),Un(x)=(−1)nUn(−x).
令 x = cos ( θ ) x=\cos(\theta) x=cos(θ), 则 T n ( x ) = cos ( n θ ) T_{n}(x) = \cos(n\theta) Tn(x)=cos(nθ)或 T n ( x ) = cos ( n arccos x ) , ( − 1 ≤ x ≤ 1 ) T_{n}(x)=\cos(n\arccos x), (-1\leq x\leq 1) Tn(x)=cos(narccosx),(−1≤x≤1) 为 n n n次Chebyshev多项式。
T n ( x ) T_{n}(x) Tn(x)在 [ − 1 , 1 ] [-1,1] [−1,1]中有 n n n个单根,分别为 x k = cos ( 2 k − 1 2 n π ) , k = 1 , 2 , ⋯ , n x_{k}=\cos(\frac{2k-1}{2n}\pi), k=1,2,\cdots,n xk=cos(2n2k−1π),k=1,2,⋯,n(或 x k = cos ( k + 1 2 ) π n , k = 0 , 1 , ⋯ , n − 1 x_{k}=\cos(k+\frac{1}{2})\frac{\pi}{n},k=0,1,\cdots,n-1 xk=cos(k+21)nπ,k=0,1,⋯,n−1).
说明:当 n n n确定的情况下, 只需令 x = cos θ , 0 ≤ θ ≤ π , − 1 ≤ x ≤ 1 x=\cos{\theta}, 0\leq \theta\leq \pi, -1\leq x\leq 1 x=cosθ,0≤θ≤π,−1≤x≤1,
θ = π / 2 n , 3 π / 2 n , ⋯ , ( 2 k − 1 ) π / 2 n , ⋯ , ( 2 n − 1 ) π / 2 n \theta = \pi/2n, 3\pi/2n,\cdots, (2k-1)\pi/2n, \cdots, (2n-1)\pi/2n θ=π/2n,3π/2n,⋯,(2k−1)π/2n,⋯,(2n−1)π/2n时, cos n θ = 0 \cos n\theta=0 cosnθ=0.
T n ( x ) T_{n}(x) Tn(x)在 [ − 1 , 1 ] [-1,1] [−1,1]中有 n + 1 n+1 n+1个极值点,为 x k = cos ( k n π ) , k = 0 , 1 , ⋯ , n x_{k}=\cos(\frac{k}{n}\pi), k=0,1,\cdots, n xk=cos(nkπ),k=0,1,⋯,n, 且对于每个极值点,其极值为 T n ( x k ) = ( − 1 ) k T_{n}(x_{k})=(-1)^{k} Tn(xk)=(−1)k.
说明: θ = 0 , π / n , 2 π / n , ⋯ , k π / n , ⋯ , π \theta =0, \pi/n, 2\pi/n, \cdots,k\pi/n, \cdots, \pi θ=0,π/n,2π/n,⋯,kπ/n,⋯,π时, cos n θ = 1 , − 1 , 1 , − 1 , ⋯ , ( − 1 ) n \cos n\theta=1,-1,1,-1,\cdots, (-1)^{n} cosnθ=1,−1,1,−1,⋯,(−1)n.
所以,在 [ − 1 , 1 ] [-1,1] [−1,1]上,Chebyshev多项式有 n n n个单根和 n + 1 n+1 n+1个极值点.
任意给定 n n n次多项式 f f f, 首项系数为 1 1 1, 那么一定有一个 x 0 ∈ [ − 1 , 1 ] x_{0}\in [-1,1] x0∈[−1,1],使得 ∣ f ( x 0 ) ∣ ≥ 1 2 n − 1 |f(x_{0})|\geq \frac{1}{2^{n-1}} ∣f(x0)∣≥2n−11.
证明: 考虑 2 n − 1 f ( x ) = F ( x ) 2^{n-1}f(x)=F(x) 2n−1f(x)=F(x),
利用反证法,若 x ∈ [ − 1 , 1 ] x\in [-1,1] x∈[−1,1], F ( x ) < 1 F(x)<1 F(x)<1,
引入 n n n次Chebyshev多项式 g ( x ) g(x) g(x),
已知 x = 1 , cos ( π / n ) , cos ( 2 π / n ) , ⋯ , cos ( k π / n ) , ⋯ , − 1 x=1,\cos(\pi/n), \cos(2\pi/n),\cdots,\cos(k\pi/n),\cdots,-1 x=1,cos(π/n),cos(2π/n),⋯,cos(kπ/n),⋯,−1这 n + 1 n+1 n+1个点上 ∣ g ( x ) ∣ = 1 |g(x)|=1 ∣g(x)∣=1,且正负不断交替。
因此, F ( x ) − g ( x ) F(x)-g(x) F(x)−g(x)是 n − 1 n-1 n−1次多项式,且在上述 n + 1 n+1 n+1个点上不断变换符号。由介值定理知,存在 y 1 , y 2 , ⋯ , y n y_{1},y_{2},\cdots,y_{n} y1,y2,⋯,yn使得 F ( y k ) = g ( y k ) F(y_{k})=g(y_{k}) F(yk)=g(yk).
对于, n − 1 n-1 n−1次多项式,有 n n n个零点,所以
F ( x ) ≡ g ( x ) F(x)\equiv g(x) F(x)≡g(x)
而Chebyshev多项式 g ( x ) g(x) g(x)在 [ − 1 , 1 ] [-1,1] [−1,1]上最大值为 1 1 1,所以与假设矛盾。
进一步地,当且仅当 F ( x ) F(x) F(x)为首1的Chebyshev多项式时, ∣ F ( x ) ∣ |F(x)| ∣F(x)∣最大值等于 1 2 n − 1 \frac{1}{2^{n-1}} 2n−11.
因此, 在区间 [ − 1 , 1 ] [-1,1] [−1,1]上所有的首1的 n n n次多项式中, 2 1 − n T n ( x ) 2^{1-n}T_{n}(x) 21−nTn(x)对零的偏差最小,即
∥ f n ( x ) ∥ ∞ ≥ ∥ 2 1 − n T n ∥ ∞ = 2 1 − n , \|f_{n}(x)\|_{\infty}\geq \|2^{1-n}T_{n}\|_{\infty}=2^{1-n}, ∥fn(x)∥∞≥∥21−nTn∥∞=21−n,
其中 f n ( x ) f_{n}(x) fn(x)是任一首1的 n n n次多项式。
(notes:本部分内容结合[3]和[4])
设 f ( x ) f(x) f(x)为定义在区间 [ − 1 , 1 ] [-1,1] [−1,1]上的函数.
并设 f ( x ) f(x) f(x)具有 ( n + 1 ) (n+1) (n+1)阶连续导数 f ( n + 1 ) ( x ) f^{(n+1)}(x) f(n+1)(x)。 给定区间 [ − 1 , 1 ] [-1,1] [−1,1]内的 ( n + 1 ) (n+1) (n+1)个互异点 x 0 , x 1 , ⋯ , x n x_{0},x_{1},\cdots,x_{n} x0,x1,⋯,xn作为插值节点,作 f ( x ) f(x) f(x)的 n n n次插值多项式:
L n ( x ) = ∑ i = 0 n f ( x i ) ∏ j = 0 , j ≠ i n x − x j x i − x j , L_{n}(x)=\sum_{i=0}^{n}f(x_{i})\prod_{j=0,j\neq i}^{n}\frac{x-x_{j}}{x_{i}-x_{j}}, Ln(x)=i=0∑nf(xi)j=0,j=i∏nxi−xjx−xj,
则插值余项为
R n ( x ) = f ( x ) − L n ( x ) = f ( n + 1 ) ( ζ ) ( n + 1 ) ! W n + 1 ( x ) , R_{n}(x) = f(x)-L_{n}(x)=\frac{f^{(n+1)}(\zeta)}{(n+1)!}W_{n+1}(x), Rn(x)=f(x)−Ln(x)=(n+1)!f(n+1)(ζ)Wn+1(x),
其中, W n + 1 ( x ) = ∏ i = 0 n ( x − x i ) , ζ ∈ ( − 1 , 1 ) W_{n+1}(x)=\prod_{i=0}^{n}(x-x_{i}),\zeta\in(-1,1) Wn+1(x)=∏i=0n(x−xi),ζ∈(−1,1)。
要使构造的插值多项式 L n ( x ) L_{n}(x) Ln(x)最优地逼近 f ( x ) f(x) f(x),也就是要使插值余项 R n ( x ) R_{n}(x) Rn(x)最小化。然而,余项 R n ( x ) R_{n}(x) Rn(x)的计算很复杂,一个替换思路是:最小化余项 ∣ R n ( x ) ∣ |R_{n}(x)| ∣Rn(x)∣的最大值,也即
min x 0 , ⋯ , x n max x ∈ [ − 1 , 1 ] ∣ R n ( x ) ∣ \min _{x_{0},\cdots,x_{n}}\max_{x\in [-1,1]}|R_{n}(x)| x0,⋯,xnminx∈[−1,1]max∣Rn(x)∣
实质上(前部分因子与 x i x_{i} xi的选择无关,是定值),也就是最小化 W n + 1 ( x ) W_{n+1}(x) Wn+1(x)的最大值,即
min x 0 , ⋯ , x n max x ∈ [ − 1 , 1 ] ∣ W n + 1 ( x ) ∣ . \min_{x_{0},\cdots,x_{n}}\max_{x\in[-1,1]}|W_{n+1}(x)|. x0,⋯,xnminx∈[−1,1]max∣Wn+1(x)∣.
由于 W n + 1 ( x ) W_{n+1}(x) Wn+1(x)是首1的 n + 1 n+1 n+1次多项式,要使在区间 [ − 1 , 1 ] [-1,1] [−1,1]上 ∥ W n + 1 ∥ \|W_{n+1}\| ∥Wn+1∥为最小,由2.1.3小节可知,插值节点应取 ( n + 1 ) (n+1) (n+1)次Chebyshev多项式 T n + 1 ( x ) T_{n+1}(x) Tn+1(x)的零点,
x k = cos ( k + 1 2 ) π n + 1 , ( k = 0 , 1 , ⋯ , n ) x_{k}=\cos(k+\frac{1}{2})\frac{\pi}{n+1}, (k=0,1,\cdots,n) xk=cos(k+21)n+1π,(k=0,1,⋯,n)
则
max − 1 ≤ x ≤ 1 ∣ f ( x ) − L n ( x ) ∣ ≤ 1 ( n + 1 ) ! max − 1 ≤ x ≤ 1 ∣ f ( n + 1 ) ( x ) ∣ ⋅ max − 1 ≤ x ≤ 1 ∣ W n + 1 ( x ) ∣ = 1 ( n + 1 ) ! max − 1 ≤ x ≤ 1 ∣ f ( n + 1 ) ( x ) ∣ ⋅ max − 1 ≤ x ≤ 1 ∣ 2 − n T n + 1 ( x ) ∣ = 2 − n ( n + 1 ) ! max − 1 ≤ x ≤ 1 ∣ f ( n + 1 ) ( x ) ∣ \begin{aligned} \max_{-1\leq x\leq 1}|f(x)-L_{n}(x)| &\leq \frac{1}{(n+1)!}\max_{-1\leq x\leq 1}|f^{(n+1)}(x)|\cdot\max_{-1\leq x\leq 1}|W_{n+1}(x)|\\ & = \frac{1}{(n+1)!}\max_{-1\leq x\leq 1}|f^{(n+1)}(x)|\cdot \max_{-1\leq x\leq 1}|2^{-n}T_{n+1}(x)|\\ &=\frac{2^{-n}}{(n+1)!}\max_{-1\leq x\leq 1}|f^{(n+1)}(x)| \end{aligned} −1≤x≤1max∣f(x)−Ln(x)∣≤(n+1)!1−1≤x≤1max∣f(n+1)(x)∣⋅−1≤x≤1max∣Wn+1(x)∣=(n+1)!1−1≤x≤1max∣f(n+1)(x)∣⋅−1≤x≤1max∣2−nTn+1(x)∣=(n+1)!2−n−1≤x≤1max∣f(n+1)(x)∣
设 f ( x ) f(x) f(x)为定义在区间 [ a , b ] [a,b] [a,b]上的函数.
可作变量代换 2
x i = a + b 2 + b − a 2 t i , ( i = 0 , 1 , ⋯ , n ) x_{i}=\frac{a+b}{2}+\frac{b-a}{2}t_{i}, (i=0,1,\cdots,n) xi=2a+b+2b−ati,(i=0,1,⋯,n)
则
max a ⩽ x ⩽ b ∣ f ( x ) − L n ( x ) ∣ = max a ⩽ x ⩽ b ∣ f ( n + 1 ) ( ξ ) ( n + 1 ) ! ∏ i = 0 n ( x − x i ) ∣ ⩽ 1 ( n + 1 ) ! max a ⩽ x ⩽ b ∣ f ( n + 1 ) ( x ) ∣ ⋅ max a ⩽ x ⩽ b ∣ ∏ i = 0 n ( x − x i ) ∣ = 1 ( n + 1 ) ! max a ⩽ x ⩽ b ∣ f ( n + 1 ) ( x ) ∣ ⋅ max − 1 ⩽ k ⩽ 1 ∣ ( b − a 2 ) n + 1 ∏ i = 0 n ( t − t i ) ∣ \begin{aligned} \max _{a \leqslant x \leqslant b} \mid & f(x)-L_{n}(x) \mid \\ &=\max _{a \leqslant x \leqslant b}\left|\frac{f^{(n+1)}(\xi)}{(n+1) !} \prod_{i=0}^{n}\left(x-x_{i}\right)\right| \\ & \leqslant \frac{1}{(n+1) !} \max _{a \leqslant x \leqslant b}\left|f^{(n+1)}(x)\right| \cdot \max _{a \leqslant x \leqslant b}\left|\prod_{i=0}^{n}\left(x-x_{i}\right)\right| \\ &=\frac{1}{(n+1) !} \max _{a \leqslant x \leqslant b}\left|f^{(n+1)}(x)\right| \cdot \max _{-1 \leqslant k \leqslant 1}\left|\left(\frac{b-a}{2}\right)^{n+1} \prod_{i=0}^{n}\left(t-t_{i}\right)\right| \end{aligned} a⩽x⩽bmax∣f(x)−Ln(x)∣=a⩽x⩽bmax∣ ∣(n+1)!f(n+1)(ξ)i=0∏n(x−xi)∣ ∣⩽(n+1)!1a⩽x⩽bmax∣ ∣f(n+1)(x)∣ ∣⋅a⩽x⩽bmax∣ ∣i=0∏n(x−xi)∣ ∣=(n+1)!1a⩽x⩽bmax∣ ∣f(n+1)(x)∣ ∣⋅−1⩽k⩽1max∣ ∣(2b−a)n+1i=0∏n(t−ti)∣ ∣
当 t 0 , t 1 , ⋯ , t n t_{0},t_{1},\cdots,t_{n} t0,t1,⋯,tn为 T n + 1 ( t ) T_{n+1}(t) Tn+1(t)的 ( n + 1 ) (n+1) (n+1)个零点,即 t i = cos ( i + 1 2 ) π n + 1 ( i = 0 , 1 , ⋯ , n ) t_{i}=\cos(i+\frac{1}{2})\frac{\pi}{n+1}(i=0,1,\cdots,n) ti=cos(i+21)n+1π(i=0,1,⋯,n)时, max − 1 ≤ x ≤ 1 ∣ ∏ i = 0 n ( t − t i ) ∣ \max_{-1\leq x\leq 1}|\prod_{i=0}^{n}(t-t_{i})| max−1≤x≤1∣∏i=0n(t−ti)∣取得最小值 2 − n 2^{-n} 2−n。因而当插值点取为
x i = a + b 2 + b − a 2 cos ( i + 1 2 ) π n + 1 ( i = 0 , 1 , ⋯ , n ) , (10) x_{i}=\frac{a+b}{2}+\frac{b-a}{2}\cos(i+\frac{1}{2})\frac{\pi}{n+1} (i=0,1,\cdots,n), \tag{10} xi=2a+b+2b−acos(i+21)n+1π(i=0,1,⋯,n),(10)
有下列余项估计式
max a ⩽ x ⩽ b ∣ f ( x ) − L n ( x ) ∣ ⩽ 1 ( n + 1 ) ! max a ⩽ x ⩽ b ∣ f ( n + 1 ) ( x ) ∣ ⋅ max − 1 ⩽ k ⩽ 1 ∣ ( b − a 2 ) n + 1 2 − n T n + 1 ( t ) ∣ = ( b − a 2 ) n + 1 ⋅ 2 − n ( n + 1 ) ! max a ⩽ x ⩽ b ∣ f ( n + 1 ) ( x ) ∣ \begin{array}{l}\max _{a \leqslant x \leqslant b}\left|f(x)-L_{n}(x)\right| \\ \quad \leqslant \frac{1}{(n+1) !} \max _{a \leqslant x \leqslant b}\left|f^{(n+1)}(x)\right| \cdot \max _{-1 \leqslant k \leqslant 1}\left|\left(\frac{b-a}{2}\right)^{n+1} 2^{-n} T_{n+1}(t)\right| \\ \quad=\left(\frac{b-a}{2}\right)^{n+1} \cdot \frac{2^{-n}}{(n+1) !} \max _{a \leqslant x \leqslant b}\left|f^{(n+1)}(x)\right|\end{array} maxa⩽x⩽b∣f(x)−Ln(x)∣⩽(n+1)!1maxa⩽x⩽b∣ ∣f(n+1)(x)∣ ∣⋅max−1⩽k⩽1∣ ∣(2b−a)n+12−nTn+1(t)∣ ∣=(2b−a)n+1⋅(n+1)!2−nmaxa⩽x⩽b∣ ∣f(n+1)(x)∣ ∣
以公式(10)中的 x 0 , x 1 , ⋯ , x n x_{0},x_{1},\cdots,x_{n} x0,x1,⋯,xn作为插值节点求得的 f ( x ) f(x) f(x)的 n n n次插值多项式 L n ( x ) L_{n}(x) Ln(x)虽不能作为 f ( x ) f(x) f(x)的 n n n次最佳一致逼近多项式,但由于误差分布比较均匀,可以作为 f ( x ) f(x) f(x)的 n n n次近似最佳一致逼近多项式。
一般正交多项式性质
初识Chebyshev多项式
参考文献:
[1]https://baike.baidu.com/item/%E5%88%87%E6%AF%94%E9%9B%AA%E5%A4%AB%E5%A4%9A%E9%A1%B9%E5%BC%8F/9390867?fr=kg_general
[2]https://wenku.baidu.com/view/d7b06e35a000a6c30c22590102020740bf1ecd53.html
[3] https://zhuanlan.zhihu.com/p/141957872
[4] 孙志忠,袁慰平,闻震初. 数值分析(第3版),东南大学出版社.
[5] Theodore J.Rivlin. Chebyshev Polynomials: From approximation theory to Algebra and number theory (Second Edition), 1990
[6] Li Hanyu, Zhu Yuanyang. Randomized block krylov space methods for trace and log-determinant estimators. BIT Numerical Mathematics (2021) 61:911–939.
[7] Walter Gautschi. Orthogonal polynomials, Quadrature, and Approximation: Computational methods and software (in Matlab)
当 x ∈ [ − 1 , 1 ] x\in [-1,1] x∈[−1,1]时,
由于 T n ( x ) = cos n θ T_{n}(x) = \cos n\theta Tn(x)=cosnθ, x = cos θ x= \cos \theta x=cosθ, 而 cos n θ = e i n θ + e − i n θ 2 \cos n\theta=\frac{e^{in\theta}+e^{-in\theta}}{2} cosnθ=2einθ+e−inθ,可验证 e i n θ = x + x 2 − 1 , e − i n θ = x − x 2 − 1 e^{in\theta}=x+\sqrt{x^{2}-1}, e^{-in\theta}=x-\sqrt{x^{2}-1} einθ=x+x2−1,e−inθ=x−x2−1,所以
T n ( x ) = cos n θ = ( x + x 2 − 1 ) n + ( x − x 2 − 1 ) n 2 , ( x ∈ [ − 1 , 1 ] ) T_{n}(x) =\cos n\theta = \frac{(x+\sqrt{x^{2}-1})^{n}+(x-\sqrt{x^{2}-1})^{n}}{2}, (x\in [-1,1]) Tn(x)=cosnθ=2(x+x2−1)n+(x−x2−1)n,(x∈[−1,1])
当 x ≥ 1 x\geq 1 x≥1时,[5]中习题1.1.2指出
T n ( x ) = cosh n t , T_{n}(x) = \cosh nt, Tn(x)=coshnt,
其中 x = cosh t , t ≥ 0 x=\cosh t, t\geq 0 x=cosht,t≥0.
由于 cosh t = e t + e − t 2 \cosh t = \frac{e^{t}+e^{-t}}{2} cosht=2et+e−t,可知
e t + e − t = 2 x e^{t}+e^{-t}=2x et+e−t=2x
求解可得
e t = x + x 2 − 1 e^{t} = x+\sqrt{x^{2}-1} et=x+x2−1
进一步可得, e − t = x − x 2 − 1 e^{-t}=x-\sqrt{x^{2}-1} e−t=x−x2−1。因此
T n ( x ) = cosh n t = ( x + x 2 − 1 ) n + ( x − x 2 − 1 ) n 2 , x ≥ 1. T_{n}(x) = \cosh nt=\frac{(x+\sqrt{x^{2}-1})^{n}+(x-\sqrt{x^{2}-1})^{n}}{2}, x\geq 1. Tn(x)=coshnt=2(x+x2−1)n+(x−x2−1)n,x≥1.
综上,可得
T n ( x ) = ( x + x 2 − 1 ) n + ( x − x 2 − 1 ) n 2 , x ≥ − 1. T_{n}(x)=\frac{(x+\sqrt{x^{2}-1})^{n}+(x-\sqrt{x^{2}-1})^{n}}{2}, x\geq -1. Tn(x)=2(x+x2−1)n+(x−x2−1)n,x≥−1. ↩︎
通过变量代换,
x = a + b 2 + b − a 2 t x=\frac{a+b}{2}+\frac{b-a}{2}t x=2a+b+2b−at
定义在区间 [ a , b ] [a,b] [a,b]上的函数 f ( x ) f(x) f(x)化为定义在区间 [ − 1 , 1 ] [-1,1] [−1,1]上的函数
g ( t ) = f ( a + b 2 + b − a 2 t ) . g(t)=f(\frac{a+b}{2}+\frac{b-a}{2}t). g(t)=f(2a+b+2b−at).
此时,找到 g ( t ) g(t) g(t)在 [ − 1 , 1 ] [-1,1] [−1,1]上的零点 t i t_{i} ti,也即找到了 f ( x ) f(x) f(x)在 [ a , b ] [a,b] [a,b]上的零点 x i x_{i} xi。 ↩︎