1
$\begingroup$

Let $N_1:=\chi_{[0,1]}$ be defined as this characteristic function and $N_n:=N_{n-1}*N_1$ then this leads to polynomials with support $[0,n]$. These splines are well-studied click for wikipedia My question is now: If you take the polynomial $N_n|_{[i,i+1]}$ and consider its unique smooth extension on $[0,n]$ which we denote $f_n,$ then I guess that $f_1,..,f_n$ are linearly independent.

Example:

For $n=2$ we get $N_2(x):=x$ on $[0,1]$ and $N_2(x):=-x+2$ on $[1,2]$ and $N_2$ is zero anywhere else. Now, it is true that $f_1(x):=x$ and $f_2(x):=-x+2$ are linearly independent as functions on $[0,2]$.

Is there a way to prove that this is true for all $n$? Due to their recursive definition, there is at least for me, no obvious way to do this.

$\endgroup$
  • $\begingroup$ What do you mean by "its unique smooth extension"? $\endgroup$ – Dirk May 19 '15 at 6:43
0
$\begingroup$

Your conjecture is true. Assume that $n$ is the smallest integer such that these polynomials are linearly dependent. Then there exist $a_0,\dots,a_{n-1}$ with $$f(x):=\sum_{i=0}^{n-1} a_i N_n(x+i)=0$$ for all $x\in [0,1]$. Let $$g(x):=\sum_{i=0}^{n-1} a_i N_{n-1}(x+i).$$ By the definition of the splines we have $$f(x)=(N_1*g)(x)=0$$ for all $x\in [0,1]$. It follows that $$\int_0^1 g(y)dy=0 \quad \text{and} \int_0^x g(y)dy=\int_1^{1+x} g(y)dy$$ for all $x\in [0,1]$. It follows that $$g(1+x)=g(x) \text{ for all } x\in [0,1]$$ and $$g^{(i)}(0)=g^{(i)}(1) \text{ for all } i\in \{0,\dots,n-3\}$$ Let $p$ be the smooth extension of $g|_{[0,1]}$. Assume that $p\neq 0$ and let $a_kx^k$ be the leading term of $p$. Then $0=g^{(k-1)}(1)-g^{(k-1)}(0)=k!a_k$ which is a contradiction. Hence $g(x)=0$ for all $x\in [0,1]$. As $N_n|_{[n-1,n]}\neq 0$ at least one of the coefficient $a_0,\dots,a_{n-2}$ is nonzero. It follows that $$\{ N_{n-1}|_{[i,i+1]}|i\in \{0,\dots,n-2\}\}$$ are linear dependent which is a contradiction to the choice of $n$ being minimal.

| cite | improve this answer | |
$\endgroup$

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.