I'm reading Interpolation and Approximation by Davis, more specifically "Best Approximation" Chapter VII.

Let $n \in \mathbb N$.

Let $C[a,b]$ denote the set of continuous real functions over $[a,b]$ and $f\in C[a,b]$

Let $\cal P_n$ denote the set of polynomials with degree less than $n$.

The author proves that $$\min_{\large P \in \cal P_n} \max_{\large x\in [a,b]} |f(x)-P(x)|$$ is attained for a unique $P_n$ which defines as "the Chebyshev approximation of degree $\leq n$".

Nevertheless he fails to mention any link between $P_n$ and the usual $\sum_{k=0}^{\lfloor n/2\rfloor} \binom{n}{2k} (x^2-1)^k x^{n-2k} $.

  • So my question is: how to prove that $P_n = \sum_{k=0}^{\lfloor n/2\rfloor} \binom{n}{2k} (X^2-1)^k X^{n-2k}$ ?

The most straightforward way would be to check that the above minimum is achieved for $\sum_{k=0}^{\lfloor n/2\rfloor} \binom{n}{2k} (X^2-1)^k X^{n-2k}$ , but I can't.


1 Answer 1


The $P_n$ mentioned by Davis is the best uniform approximation of a given function $f$, so it depends on $f$. The polynomial you mentioned is the plain ordinary Chebyshev polynomial, and is independent of any function $f$. So, they can't be the same thing.

In fact, though, one is a special case of the other. One definition of the Chebyshev polynomial is that it's the polynomial with minimal uniform norm. In other words, it's a polynomial that is a best uniform approximation of the zero function. So, if you go through Davis' argument with $f=0$, then the $P_n$ that you get should be the usual Chebyshev polynomial.

  • $\begingroup$ Yes you're totally right. I was out of my mind the whole time. $\endgroup$ May 9, 2014 at 15:02
  • 1
    $\begingroup$ Wouldnt $P_n = 0?$ be the best appx for f=0? $\endgroup$
    – JustANoob
    Mar 11, 2018 at 1:17

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.