I am trying to grasp the idea of Taylor Remainder Theorem. I want to know the way I understand it is right or wrong.

Like in the linear approximation

it is $$f(x) \approx f(c) + f'(c)(x-c)$$ or $$f(x) = f(c) + f'(c)(x-c) + f''(\zeta)(x-c)$$

which is in the picture below enter image description here

The error is exactly $$f''(\zeta)(x-c)$$

Then I thought of a case with a lower order of derivative which is

$$f(x) = f(c) + f'(\zeta)(x-c)$$

enter image description here

The error is exactly $$f'(\zeta)(x-c)$$

Then I noticed as there are more terms the order of derivatives for error gets higher which leads to the equation $$\frac{f^{n+1}(\zeta)}{(n+1)!}(x-a)^{n+1}$$

Is this the right way of understanding the theorem?


If $f$ is sufficiently often differentiable in the neighborhood of the point $c\in{\mathbb R}$ then for each $n\geq0$ its $n^{\rm th}$ Taylor polynomial $j_c^nf$ is defined as follows: $$j_c^nf(x):=\sum_{k=0}^n{f^{(k)}(c)\over k!}(x-c)^k\ .$$ $\bigl($Note: Sometimes the increment variable $X:=x-c$ is used as variable for $j_c^nf$. One then writes $$j_c^nf(X):=\sum_{k=0}^n{f^{(k)}(c)\over k!}X^k\ .\bigr)$$ Given $x$ the value of such a polynomial can be computed exactly in finitely many steps.

Why should we introduce this polynomial? That's where "Taylor's theorem with remainder" comes in. It turns out that when $|x-c|$ is small this polynomial gives a good approximation to the true value of $f$ at $x$: $$f(x)\approx j_c^nf (x)\quad\bigl(|x-c|\ll1)\ .$$ Now "good approximation" is just a colloquial description. We want error bounds! There are various ways to quantify the error $$R_n(x):=f(x)-j_c^nf(x)\ .$$ One of them reads as follows: There is a point $\xi$ between $c$ and $x$ such that $$R_n(x)={f^{(n+1)}(\xi)\over (n+1)!}(x-c)^{n+1}\ .\tag{1}$$ The formula $(1)$ only is of use if you have simple control over the values of $f^{(n+1)}$. This is, e.g., the case when $f=\sin$. But don't think the formula $$f(x)=j_c^nf(x)+R_n(x)$$ allows you to compute the value of some special $f(x)$ exactly in a "finitary", albeit very complicated way.

| cite | improve this answer | |

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.