-3
$\begingroup$

I'm interested in this function

$$ h(m,n) = \operatorname{sgn}\Bigl(\lim_{w \to \infty}{\sum_{x_1=1}^{w}\dotsi\sum_{x_u=1}^{w}\dfrac{1}{1 + wp^2(m,n,x_1,\dotsc,x_u)}}\Bigr) $$

where $p$ is a polynomial in $u+2$ variables and integer coefficients. Can we simplify it to $$ h(m,n) = \operatorname{sgn}\Bigl(\dfrac{f(m,n)}{g(m,n)}\Bigr) $$

without the limit?

$\endgroup$
11
  • $\begingroup$ What’s $p$? Is $u$ given? Why the parenthesis closes there? $\endgroup$ Sep 10, 2022 at 6:28
  • $\begingroup$ @AlessandroDellaCorte $p$ is a polynomial with $u$ unknown and a specific degree $\endgroup$
    – raoof
    Sep 10, 2022 at 6:33
  • $\begingroup$ So $m,n,w,u,x_1,\dots,x_u$ are all non-negative integers and $\epsilon$ a real number, aren't they? And is $p$ a polynomial with real coefficients? $\endgroup$ Sep 10, 2022 at 7:41
  • 1
    $\begingroup$ You are asking for the sign of a sum of positive values: and you want to know if at the limit it is zero or 1, is that correct? Basically you are asking if dominated conevergence applies? Well, if the sums are finite for $\epsilon=1$, since the term being summed is monotonous in $\epsilon$, the answer is yes, it is zero, thank you Lebesgue. $\endgroup$
    – username
    Sep 10, 2022 at 8:03
  • 1
    $\begingroup$ By editing the question to remove the mention of $\epsilon$, you have made the existing answer invalid. $\endgroup$
    – JRN
    Oct 30, 2022 at 8:10

1 Answer 1

3
$\begingroup$

Assuming you mean $\epsilon$ is a real positive number and $p$ is a real-valued function in $2+u$ variables, you want a limit of a sum of non-negative numbers, $$\lim_{\epsilon\to0}\sum_{x\in\mathbb N_+^u}\frac{\epsilon^2}{\epsilon^2+p^2(n,m,x)}.$$ For $\epsilon\searrow 0$, each coefficient converges decreasing either to $1$ or $0$, according whether $p(n,m,x)=0$ or not. Therefore, if the sum is finite for at least one $\epsilon$, by Beppo Levi the limit is just the number of non-negative integer zeros $x\in\mathbb N_+^u$ of $p(n,m,\cdot\}$; and if $\text{sgn}$ is the usual signum function, the final expression is $1$ or $0$ according whether $p(n,m,\cdot)$ has a zero $x\in\mathbb N_+^u$ or not.

$\endgroup$
5
  • $\begingroup$ So everything reduces to checking if the sum is finite for at least one $\epsilon>0$ (hence for all, since each term is concave w.r.to $\epsilon$). If e.g. $p$ is a constant, the sum is infinite for all $\epsilon>0$. $\endgroup$ Sep 10, 2022 at 8:37
  • $\begingroup$ $p$ is a integer valued function. what I have in mind is whether it can be simplified to $f(m,n)/g(m,n)$ without the limit so my function could have some undefined values $\endgroup$
    – raoof
    Sep 10, 2022 at 8:53
  • $\begingroup$ @PietroMajer You say Beppo-Levi, I say Lebesgue : Potayto/Potahto... : ) $\endgroup$
    – username
    Sep 10, 2022 at 11:31
  • $\begingroup$ I really mean beppi levi (monotone convergence; no need of finiteness if it was increasing); though lebesgue would also do (dominated convergence) $\endgroup$ Sep 10, 2022 at 12:00
  • $\begingroup$ The question has been changed; $\epsilon$ is now no longer mentioned. Your answer has become invalid. $\endgroup$
    – JRN
    Oct 30, 2022 at 8:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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