0
$\begingroup$

Let $n\in\mathbb{N}$ such that $n\geq 2$ and let $0<r<1$ be a real number. We wish to solve the following problem.

$$\min_{(t,(z_j)_{j=2}^n) \in \mathbb{R}\times \mathbb{Z}^{n-1}} t$$

subject to :

  • $\left\lvert\left(\frac{\ln(j)}{2\pi}\right)t - z_j\right\rvert \leq \frac{r}{2\pi} \quad \text{ for } 2 \leq j \leq n$.

  • $z_j\geq 0 \quad \text{ for } 2 \leq j \leq n$.

  • $\sum_{j=2}^n z_j \geq 1$.

  • $t \geq 0$.

Note that $(z_j)_{j=2}^n$ is a non-zero vector of natural numbers. Suppose, we write the optimal value $t^*$ as a function of $n$ and $r$ i.e. say $t^* = t^*(n,r)$.

  1. How does $t^*(n,r)$ depend on $n$ and $r$?

  2. Is it independent of $n$ and only dependent on $r$?

  3. Is it polynomial or exponential in $n$?

  4. Could we derive a good upper bound for $t^*(n,r)$?

$\endgroup$
1
  • $\begingroup$ @RobPratt Thanks! These graph plots seem interesting. It seems that no one was able to formally prove anything regarding the above problem about the exact nature of dependence of the optimal value on n and r. One may think of the upper bound problem in 4 and try to prove results about asymptotics on n . Please also see the following mathoverflow post regarding a similar puzzle that might provide insight : mathoverflow.net/questions/454856/linear-program-optimal-value. One may possibly define polynomials as a linear combination of monomials with rational powers. $\endgroup$ Commented Oct 26, 2023 at 22:30

1 Answer 1

0
$\begingroup$

It certainly depends on both $n$ and $r$. Here are plots of $t^*(n,r)$ for $n\in\{2,3,4,5\}$ and $r\in\{0.01,0.02,\dots,0.99\}$, obtained via mixed integer linear programming.

$$n=2$$

enter image description here


$$n=3$$

enter image description here


$$n=4$$

enter image description here


$$n=5$$

enter image description here

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .