C.5 El comportamiento de error del método secante
- Page ID
- 118164
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\( \newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\)
( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\id}{\mathrm{id}}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\kernel}{\mathrm{null}\,}\)
\( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\)
\( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\)
\( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\AA}{\unicode[.8,0]{x212B}}\)
\( \newcommand{\vectorA}[1]{\vec{#1}} % arrow\)
\( \newcommand{\vectorAt}[1]{\vec{\text{#1}}} % arrow\)
\( \newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vectorC}[1]{\textbf{#1}} \)
\( \newcommand{\vectorD}[1]{\overrightarrow{#1}} \)
\( \newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}} \)
\( \newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}} \)
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\(\newcommand{\avec}{\mathbf a}\) \(\newcommand{\bvec}{\mathbf b}\) \(\newcommand{\cvec}{\mathbf c}\) \(\newcommand{\dvec}{\mathbf d}\) \(\newcommand{\dtil}{\widetilde{\mathbf d}}\) \(\newcommand{\evec}{\mathbf e}\) \(\newcommand{\fvec}{\mathbf f}\) \(\newcommand{\nvec}{\mathbf n}\) \(\newcommand{\pvec}{\mathbf p}\) \(\newcommand{\qvec}{\mathbf q}\) \(\newcommand{\svec}{\mathbf s}\) \(\newcommand{\tvec}{\mathbf t}\) \(\newcommand{\uvec}{\mathbf u}\) \(\newcommand{\vvec}{\mathbf v}\) \(\newcommand{\wvec}{\mathbf w}\) \(\newcommand{\xvec}{\mathbf x}\) \(\newcommand{\yvec}{\mathbf y}\) \(\newcommand{\zvec}{\mathbf z}\) \(\newcommand{\rvec}{\mathbf r}\) \(\newcommand{\mvec}{\mathbf m}\) \(\newcommand{\zerovec}{\mathbf 0}\) \(\newcommand{\onevec}{\mathbf 1}\) \(\newcommand{\real}{\mathbb R}\) \(\newcommand{\twovec}[2]{\left[\begin{array}{r}#1 \\ #2 \end{array}\right]}\) \(\newcommand{\ctwovec}[2]{\left[\begin{array}{c}#1 \\ #2 \end{array}\right]}\) \(\newcommand{\threevec}[3]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \end{array}\right]}\) \(\newcommand{\cthreevec}[3]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \end{array}\right]}\) \(\newcommand{\fourvec}[4]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}\) \(\newcommand{\cfourvec}[4]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}\) \(\newcommand{\fivevec}[5]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}\) \(\newcommand{\cfivevec}[5]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}\) \(\newcommand{\mattwo}[4]{\left[\begin{array}{rr}#1 \amp #2 \\ #3 \amp #4 \\ \end{array}\right]}\) \(\newcommand{\laspan}[1]{\text{Span}\{#1\}}\) \(\newcommand{\bcal}{\cal B}\) \(\newcommand{\ccal}{\cal C}\) \(\newcommand{\scal}{\cal S}\) \(\newcommand{\wcal}{\cal W}\) \(\newcommand{\ecal}{\cal E}\) \(\newcommand{\coords}[2]{\left\{#1\right\}_{#2}}\) \(\newcommand{\gray}[1]{\color{gray}{#1}}\) \(\newcommand{\lgray}[1]{\color{lightgray}{#1}}\) \(\newcommand{\rank}{\operatorname{rank}}\) \(\newcommand{\row}{\text{Row}}\) \(\newcommand{\col}{\text{Col}}\) \(\renewcommand{\row}{\text{Row}}\) \(\newcommand{\nul}{\text{Nul}}\) \(\newcommand{\var}{\text{Var}}\) \(\newcommand{\corr}{\text{corr}}\) \(\newcommand{\len}[1]{\left|#1\right|}\) \(\newcommand{\bbar}{\overline{\bvec}}\) \(\newcommand{\bhat}{\widehat{\bvec}}\) \(\newcommand{\bperp}{\bvec^\perp}\) \(\newcommand{\xhat}{\widehat{\xvec}}\) \(\newcommand{\vhat}{\widehat{\vvec}}\) \(\newcommand{\uhat}{\widehat{\uvec}}\) \(\newcommand{\what}{\widehat{\wvec}}\) \(\newcommand{\Sighat}{\widehat{\Sigma}}\) \(\newcommand{\lt}{<}\) \(\newcommand{\gt}{>}\) \(\newcommand{\amp}{&}\) \(\definecolor{fillinmathshade}{gray}{0.9}\)Dejar\(f(x)\) tener dos derivadas continuas, y dejar\(r\) ser cualquier solución de Ahora\(f(x)=0\text{.}\) vamos a obtener un manejo bastante bueno sobre el comportamiento de error del método secante cerca\(r\text{.}\)
Denote por\(\tilde\varepsilon_n=x_n-r\) el error (firmado) en\(x_n\) y por\(\varepsilon_n=|x_n-r|\) el error (absoluto) en\(x_n\text{.}\) Entonces,\(x_n =r+\tilde\varepsilon_n\text{,}\) y, por (C.4.1),
\ begin {align*}\ tilde\ varepsilon_ {n+1} & =\ frac {x_ {n-1} f (x_n) - x_n f (x_ {n-1})} {f (x_n) -f (x_ {n-1})} -r\\ & =\ frac {[r+\ tilde\ varepsilon_ {n-1}] f (x_n) - [r+\ tilde\ varepsilon_n] f (x_ {n-1})} {f (x_n) -f (x_ {n-1})} -r\\ & =\ frac {\ tilde\ varepsilon_ {n-1} f (x_n) -\ tilde\ varepsilon_ _n f (x_ {n-1})} {f (x_n) - f (x_ {n-1})}\ final {alinear*}
Por la expansión de Taylor (3.4.32) y el teorema del valor medio (Teorema 2.13.5),
\ begin {align*} f (x_n) & = f (r) + f' (r)\ tilde\ varepsilon_n +\ frac {1} {2} f "(c_1)\ tilde\ varepsilon_n^2\\ & = f' (r)\ tilde\ varepsilon_n +\ frac {1} {2} f" (c_1)\ tilde\ varepsilon_n^2\\ f (x_n) -f (x_ {n-1}) & = f' (c_2) [x_n-x_ {n-1}]\\ & = f' (c_2) [\ tilde\ varepsilon_n-\ tilde\ varepsilon_ {n-1}]\ end {alinear*}
para algunos\(c_1\) entre\(r\) y\(x_n\) y algunos\(c_2\) entre\(x_{n-1}\) y\(x_n\text{.}\) Entonces, para\(x_{n-1}\) y\(x_n\) cerca\(r\text{,}\)\(c_1\) y\(c_2\) también tienen que estar cerca\(r\) y
\ begin {align*} f (x_n) &\ approx f' (r)\ tilde\ varepsilon_n +\ frac {1} {2} f "(r)\ tilde\ varepsilon_n^2\\ f (x_ {n-1}) &\ approx f' (r)\ tilde\ varepsilon_ {n-1} +\ frac {1} {2} f" (r)\ tilde\ varepsilon_ {n-1} ^2\\ f (x_n) -f (x_ {n-1}) &\ approx f' (r) [\ tilde\ varepsilon_n-\ tilde\ varepsilon_ {n-1}]\ end {align*}
y
\ begin {alinear*}\ tilde\ varepsilon_ {n+1} & =\ frac {\ tilde\ varepsilon_ {n-1} f (x_n) -\ tilde\ varepsilon_n f (x_ {n-1})} {f (x_n) -f (x_ {n-1})}\\ &\ aprox {\ tilde\ varepsilon_ {n-1} [f' (r)\ tilde\ varepsilon_n +\ frac {1} {2} f "(r)\ tilde\ varepsilon_n^2] -\ tilde\ varepsilon_n [f' (r)\ tilde\ varepsilon_ {n-1} +\ frac {1} {2} f "(r)\ tilde\ varepsilon_ {n-1} ^2]} {f' (r) [\ tilde\ varepsilon_n-\ tilde\ varepsilon_ {n-1}]}\\ & =\ frac {\ frac {1} {2}\ tilde\ varepsilon_ {n-1}\ tilde\ varepsilon_nf" (r) [\ tilde\ varepsilon_n-\ tilde\ varepsilon_ {n-1}]} {f' (r) [\ tilde\ varepsilon_n-\ tilde\ varepsilon_ {n-1}]}\\ & =\ frac {f "(r)} {2f' (r)}\ tilde\ varepsilon_ {n-1}\ tilde\ varepsilon_n\ final {alinear*}
Tomando valores absolutos, tenemos
\[ \varepsilon_{n+1}\approx K \varepsilon_{n-1}\varepsilon_n\qquad \text{with }K = \left|\frac{f''(r)} {2f'(r)} \right| \tag{E7} \nonumber \]
Hemos visto que el método de Newton obedece a una fórmula similar — (E3) dice que, cuando\(x_n\) está cerca el método de\(r\text{,}\) Newton obedece\(\varepsilon_{n+1}\approx K\varepsilon_n^2\text{,}\) también con\(K = \left|\frac{f''(r)} {2f'(r)} \right|\text{.}\) Como veremos ahora, el cambio de\(\varepsilon_n^2\text{,}\) in\(\varepsilon_{n+1}\approx K\varepsilon_n^2\text{,}\) a\(\varepsilon_{n-1}\varepsilon_n\text{,}\) in\(\varepsilon_{n+1}\approx K\varepsilon_{n-1}\varepsilon_n\text{,}\) sí tiene un impacto sustancial en el comportamiento de \(\varepsilon_n\)para grandes\(n\text{.}\)
Para ver el\(n\) comportamiento grande, ahora iteramos (E7). Las fórmulas se verán más simples si multiplicamos (E7) por\(K\) y escribimos\(\delta_n=K\varepsilon_n\text{.}\) Entonces (E7) se convierte\(\delta_{n+1}\approx\delta_{n-1}\delta_n\) (y hemos eliminado\(K\)). Las primeras iteraciones son
\ begin {alignat*} {2}\ delta_2& & &\ approx\ delta_0\ delta_1\\ delta_3&\ approx\ delta_1\ delta_2 & &\ approx\ delta_0\ delta_1^2\\ delta_4&\ approx\ delta_2\ delta_3 &\ approx\ delta_0^2\ delta_1^3\\\ delta_5&\ approx\ delta_3\ delta_4 & &\ approx\ delta_0^3\ delta_ 1^5\\ delta_6&\ approx\ delta_4\ delta_5 & &\ approx\ delta_0^5\ delta_1^8\\ delta_7&\ approx\ delta_5\ delta_6 & &\ approx\ delta_0^8\ delta_1^ {13}\ end {alignat*}
Observe que cada\(\delta_n\) es de la forma\(\delta_0^{\alpha_n}\delta_1^{\beta_n}\text{.}\) Sustituir\(\delta_n=\delta_0^{\alpha_n}\delta_1^{\beta_n}\) en\(\delta_{n+1}\approx\delta_{n-1}\delta_n\) da
\[ \delta_0^{\alpha_{n+1}}\delta_1^{\beta_{n+1}} \approx \delta_0^{\alpha_{n-1}}\delta_1^{\beta_{n-1}} \delta_0^{\alpha_n}\delta_1^{\beta_n} \nonumber \]
y tenemos
\[ \alpha_{n+1}=\alpha_{n-1}+\alpha_{n} \qquad \beta_{n+1}=\beta_{n-1}+\beta_{n} \tag{E8} \nonumber \]
La regla de recursión en (E8) es famosa 1. La secuencia de Fibonacci 2 (que es\(0\text{,}\)\(1\text{,}\)\(1\text{,}\)\(2\text{,}\)\(3\text{,}\)\(5\text{,}\)\(8\text{,}\)\(13\text{,}\)\(\cdots\)), se define por
\ begin {align*} F_0& =0\\ F_1& =1\\ F_n& =F_ {n-1} +F_ {n-2}\ qquad\ text {for} n>1\ end {align*}
Entonces, para\(n\ge 2\text{,}\)\(\alpha_n = F_{n-1}\) y\(\beta_n=F_n\) y
\[ \delta_n \approx \delta_0^{\alpha_n}\delta_1^{\beta_n} = \delta_0^{F_{n-1}}\delta_1^{F_n} \nonumber \]
Una de las propiedades conocidas de la secuencia de Fibonacci es que, para\(n\text{,}\)
\[ F_n\approx\frac{\varphi^n}{\sqrt{5}}\qquad\text{where } \varphi=\frac{1+\sqrt{5}}{2} \approx 1.61803 \nonumber \]
Esta\(\varphi\) es la proporción áurea 3. Entonces, para grandes\(n\text{,}\)
\ begin {alinear*} K\ varepsilon_n & =\ delta_n\ approx\ delta_0^ {F_ {n-1}}\ delta_1^ {f_n}\ approx\ delta_0^ {\ frac {\ varphi^ {n-1}} {\ sqrt {5}}\ delta_1^ {\ frac {\ varphi^ {\ varphi^ {\ varphi^ {\ sqrt {5}}\ delta_1^ {\ frac {\ ^n} {\ sqrt {5}}} =\ delta_0^ {\ frac {1} {\ sqrt {5}\ varphi}\ veces\ varphi^n}\ delta_1^ {\ frac {1} {\ sqrt {5}}\ veces\ varphi^n}\\ & = d^ {\ varphi^n}\ qquad\ text { donde}\ quad d=\ delta_0^ {\ frac {1} {\ sqrt {5}\,\ varphi}}\ delta_1^ {\ frac {1} {\ sqrt {5}}}\\ &\ aprox d^ {1.6^n}\ end {align*}
Suponiendo que\(0\lt \delta_0=K\varepsilon_0\lt 1\) y\(0\lt \delta_1=K\varepsilon_1\lt 1\text{,}\) vamos a tener\(0\lt d\lt 1\text{.}\)
A modo de contraste, para el método de Newton, para grandes\(n\text{,}\)
\ begin {reunir*} K\ varepsilon_n\ aprox d^ {2^n}\ qquad\ text {donde}\ quad d= (K\ varepsilon_1) ^ {1/2}\ end {reunir*}
Como\(2^n\) crece bastante más rápido que\(1.6^n\) (por ejemplo, cuando n=5,\(2^n=32\) y cuándo\(1.6^n=10.5\text{,}\)\(n=10\text{,}\)\(2^n=1024\) y\(1.6^n=110\)) el método de Newton se casa en la raíz bastante más rápido que el método secante, asumiendo que comienzas razonablemente cerca de la raíz.