11.08.2013 Views

Untitled - Cdm.unimo.it

Untitled - Cdm.unimo.it

Untitled - Cdm.unimo.it

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Results in Approximation Theory 97<br />

Proof - A polynomial ψ ∈ Pn can be wr<strong>it</strong>ten in the form ψ = n<br />

k=0 dkuk, for some<br />

real coefficients dk, 0 ≤ k ≤ n. Minimizing f −ψ L 2 w (I), or equivalently f −ψ 2 L 2 w (I),<br />

requires the derivatives<br />

(6.2.3)<br />

+<br />

n<br />

k=0<br />

∂<br />

∂dj<br />

f − ψ 2 L2 <br />

∂<br />

w (I) = f<br />

∂dj<br />

2 L2 w<br />

d 2 kuk 2 L 2 w (I)<br />

<br />

(I) − 2<br />

n<br />

k=0<br />

dk(f,uk) L 2 w (I)<br />

= −2 (f,uj) L 2 w (I) + 2dj uj 2 L 2 w (I),<br />

0 ≤ j ≤ n.<br />

We deduce that the unique minimum is attained when dj = cj, 0 ≤ j ≤ n, where the<br />

cj’s are the Fourier coefficients of f (see (6.2.2)). This ends the proof.<br />

In short, we can wr<strong>it</strong>e<br />

(6.2.4) f − Πw,nf L 2 w (I) = inf<br />

ψ∈Pn<br />

f − ψ L 2 w (I).<br />

Another interesting characterization is given in the following theorem.<br />

Theorem 6.2.2 - For any f ∈ L 2 w(I) and n ∈ N, we have<br />

(6.2.5)<br />

<br />

I<br />

(f − Πw,nf)φw dx = 0, ∀φ ∈ Pn.<br />

Proof - We fix φ ∈ Pn and define G : R → R by<br />

G(ν) := f − Πw,nf + νφ 2 L 2 w (I).<br />

We know from theorem 6.2.1 that ν = 0 is a minimum for G. Therefore<br />

(6.2.6) G ′ (ν) = 2<br />

<br />

(f − Πw,nf)φw dx + 2νφ<br />

I<br />

2 L2 w (I), ∀ν ∈ R.<br />

Imposing that G ′ (0) = 0, we get (6.2.5).

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!