Searcharxiv⌕ Search

arXiv subjects

Oriol Solé Pi

Publications and source records attributed to Oriol Solé Pi.

2 recordsLinked to original sources

Covering and packing with homothets of limited capacity

This work revolves around the two following questions: Given a convex body $C\subset\mathbb{R}^d$, a positive integer $k$ and a finite set $S\subset\mathbb{R}^d$ (or a finite Borel measure $μ$ on $\mathbb{R}^d$), how many homothets of $C$ are required to cover $S$ if no homothet is allowed to cover more than $k$ points of $S$ (or have measure larger than $k$)? How many homothets of $C$ can be packed if each of them must cover at least $k$ points of $S$ (or have measure at least $k$)? We prove that, so long as $S$ is not too degenerate, the answer to both questions is $Θ_d(\frac{|S|}{k})$, where the hidden constant is independent of $d$. This is optimal up to a multiplicative constant. Analogous results hold in the case of measures. Then we introduce a generalization of the standard covering and packing densities of a convex body $C$ to Borel measure spaces in $\mathbb{R}^d$ and, using the aforementioned bounds, we show that they are bounded from above and below, respectively, by functions of $d$. As an intermediate result, we give a simple proof the existence of weak $ε$-nets of size $O(\frac{1}ε)$ for the range space induced by all homothets of $C$. Following some recent work in discrete geometry, we investigate the case $d=k=2$ in greater detail. We also provide polynomial time algorithms for constructing a packing/covering exhibiting the $Θ_d(\frac{|S|}{k})$ bound mentioned above in the case that $C$ is an Euclidean ball. Finally, it is shown that if $C$ is a square then it is NP-hard to decide whether $S$ can be covered using $\frac{|S|}{4}$ squares containing $4$ points each.

math.CO↗

Pair crossing number, cutwidth, and good drawings on arbitrary point sets

Determining whether there exists a graph such that its crossing number and pair crossing number are distinct is an important open problem in geometric graph theory. We show that $\textit{cr}(G)=O(\mathop{\mathrm{pcr}}(G)^{3/2})$ for every graph $G$, this improves the previous best bound by a logarithmic factor. Answering a question of Pach and Tóth, we prove that the bisection width (and, in fact, the cutwidth as well) of a graph $G$ with degree sequence $d_1,d_2,\dots,d_n$ satisfies $\mathop{\mathrm{bw}}(G)=O\big(\sqrt{\mathop{\mathrm{pcr}}(G)+\sum_{k=1}^n d_k^2}\big)$. Then we show that there is a constant $C\geq 1$ such that the following holds: For any graph $G$ of order $n$ and any set $S$ of at least $n^C$ points in general position on the plane, $G$ admits a straight-line drawing which maps the vertices to points of $S$ and has no more than $O\left(\log n\cdot\left(\mathop{\mathrm{pcr}}(G)+\sum_{k=1}^n d_k^2\right)\right)$ crossings. Our proofs rely on a modified version of a separator theorem for string graphs by Lee, which might be of independent interest.

math.CO↗