10
$\begingroup$

I have approximated $\sin(x)$ and $\cos (x)$ using the Taylor Series (Maclaurin Series) with the following results:

$$f(x)=f(0)+\frac{f^{(1)}(0)}{1!}(x-0)+\frac{f^{(2)}(0)}{2!}(x-0)^2+\frac{f^{(3)}(0)}{3!}(x-0)^3+\cdots$$ $$\begin{align}\implies \sin(x)&=x-\frac{x^3}{3!}+\frac{x^5}{5!}-\frac{x^7}{7!}+\frac{x^9}{9!}-\cdots\\&=\sum_{n=0}^{\infty} \frac{(-1)^n}{(2n+1)!}x^{2n+1} \end{align}$$

$$f(x)=f(a)+\frac{f^{(1)}(a)}{1!}(x-a)+\frac{f^{(2)}(a)}{2!}(x-a)^2+\frac{f^{(3)}(a)}{3!}(x-a)^3+\cdots$$ $$\begin{align} \implies \cos(x)&=1-\frac{x^2}{2!}+\frac{x^4}{4!}-\frac{x^6}{6!}+\frac{x^8}{8!}-\frac{x^10}{10!}+\cdots \\&=\sum_{n=0}^{\infty} \frac{(-1)^n}{(2n)!}x^{2n} \end{align}$$

How can I use Chebyshev Polynomials to approximate $\sin(x)$ and $\cos(x)$ within the interval $[−π,π]$?

Thanks in advance!

$\endgroup$
  • $\begingroup$ It would be better to rephrase the question in more specific terms, like: "How to compute the Fourier-Chebyshev expansion of $\sin(x)$ and $\cos(x)$ over $[-1,1]$?" - and add your attempts. The link is quite irrelevant, you may assume we know how to approximate an exponential through Chebyshev polynomials. $\endgroup$ – Jack D'Aurizio Jul 1 '15 at 12:26
  • $\begingroup$ Thanks! I've modified my question. $\endgroup$ – David Lin Jul 2 '15 at 2:23
10
$\begingroup$

Assume that we want to approximate $\cos(x)$ over $I=[-1,1]$. The Taylor series gives: $$ \cos(x) = \sum_{n=0}^{+\infty}\frac{(-1)^n}{(2n)!} x^{2n}\tag{1} $$ while the Fourier-Chebyshev series is given by: $$ \cos(x) = J_0(1) + 2\sum_{n\geq 1}(-1)^n J_{2n}(1) T_{2n}(x)\tag{2}$$ where: $$ \frac{2}{\pi}\int_{-1}^{1}\frac{\cos(x) T_n(x)}{\sqrt{1-x^2}}\,dx = \frac{2}{\pi}\int_{-\pi/2}^{\pi/2}\cos(\cos x)\cos(nx)\,dx\tag{3}$$ is a Bessel function of the first kind and $T_{n}(x)$ is a Chebyshev polynomial of the first kind. Since both $x^n$ and $T_n(x)$ are bounded by $1$ on $I$, the accuracy of the approximation just depends on how fast $J_{2n}(1)$ decays to zero. Since: $$ J_{2n}(1) = \sum_{l\geq 0}\frac{(-1)^l}{4^{l+n}(2n+l)!},\qquad \left|J_{2n}(1)\right|\approx\frac{1}{4^n(2n)!}\tag{4}$$ we have that the Taylor approximation pointwise outperforms $(2)$ in the region $|x|\leq\frac{1}{2}$, while the $L^2$-approximation $(2)$ is more accurate than $(1)$ near the endpoints of $I$, and in uniform terms. However, its coefficients are less trivial to compute.

It is also interesting to mention that, by expanding $T_{2n}(x)$ and equating the coefficients of $x^{2m}$ in $(1)$ and $(2)$ we get an interesting identity about the Bessel function of the first kind. The same happens if we integrate the square of both sides of $(2)$, multiplied by $\frac{1}{\sqrt{1-x^2}}$, over $I$.

Here we have a graph of the approximation error for $n=5$. The Taylor series gives the blue graph, the Fourier-Chebyshev series gives the red graph.

$\hspace1in$enter image description here

This is the same graph over the subinterval $\left[-\frac{2}{3},\frac{2}{3}\right]$:

$\hspace1in$enter image description here

$\endgroup$
  • $\begingroup$ Thanks! What is the difference between Chebyshev Polynomials and the Fourier-Chebyshev series? $\endgroup$ – David Lin Jul 1 '15 at 9:27
  • $\begingroup$ Also, is there a way to approximate Sine and Cosine using Chebyshev Polynomials which have results similar to caig.cs.nctu.edu.tw/course/NM07S/slides/chap4_1.pdf Page 9? $\endgroup$ – David Lin Jul 1 '15 at 11:36
  • $\begingroup$ @DavidLin: Chebyshev Polynomials : Fourier-Chebyshev series = $\sin(nx)$ : Fourier series. Chebyshev polynomials are a complete base of $L^2(-1,1)$ with respect to the inner product $\int_{-1}^{1} \frac{f\cdot g}{\sqrt{1-x^2}}\,dx$, you should know this. I do not understand your second question: are you asking me to write $x^n$ in terms of Chebyshev polynomials or the opposite? $\endgroup$ – Jack D'Aurizio Jul 1 '15 at 12:22
  • $\begingroup$ Thanks again! My second question is asking if there is a way to express sin(x) and cos(x) each in one simple line of infinite Chebyshev Series. Sorry, I am only a highschool student so I don't really quite understand 70% of the high maths in your answer LOL. Thanks again for answering and helping out! $\endgroup$ – David Lin Jul 2 '15 at 2:33
  • $\begingroup$ @DavidLin: my $(2)$ gives $\cos(x)$ in terms of Chebyshev polynomials. To have a similar expansion for $\sin(x)$, too, just differentiate both sides of $(2)$. $\endgroup$ – Jack D'Aurizio Jul 2 '15 at 11:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.