Searcharxiv⌕ Search

arXiv subjects

Václav Rozhoň

Publications and source records attributed to Václav Rozhoň.

40 records · Page 3Linked to original sources

Relating the cut distance and the weak* topology for graphons

The theory of graphons is ultimately connected with the so-called cut norm. In this paper, we approach the cut norm topology via the weak* topology (when considering a predual of $L^{1}$-functions). We prove that a sequence $W_1,W_2,W_3,\ldots$ of graphons converges in the cut distance if and only if we have equality of the sets of weak* accumulation points and of weak* limit points of all sequences of graphons $W_1',W_2',W_3',\ldots$ that are weakly isomorphic to $W_1,W_2,W_3,\ldots$. We further give a short descriptive set theoretic argument that each sequence of graphons contains a subsequence with the property above. This in particular provides an alternative proof of the theorem of Lovász and Szegedy about compactness of the space of graphons. We connect these results to "multiway cut" characterization of cut distance convergence from [Ann. of Math. (2) 176 (2012), no. 1, 151-219]. These results are more naturally phrased in the Vietoris hyperspace $K$ over graphons with the weak* topology. We show that graphons with the cut distance topology are homeomorphic to a closed subset of $K$, and deduce several consequences of this fact. From these concepts a new order on the space of graphons emerges. This order allows to compare how structured two graphons are. We establish basic properties of this "structurdness order".

math.CO↗

k-means++: few more steps yield constant approximation

The k-means++ algorithm of Arthur and Vassilvitskii (SODA 2007) is a state-of-the-art algorithm for solving the k-means clustering problem and is known to give an O(log k)-approximation in expectation. Recently, Lattanzi and Sohler (ICML 2019) proposed augmenting k-means++ with O(k log log k) local search steps to yield a constant approximation (in expectation) to the k-means clustering problem. In this paper, we improve their analysis to show that, for any arbitrarily small constant $\eps > 0$, with only $\eps k$ additional local search steps, one can achieve a constant approximation guarantee (with high probability in k), resolving an open problem in their paper.

cs.DS↗

A local approach to the Erdős-Sós conjecture

A famous conjecture of Erdős and Sós states that every graph with average degree more than $k - 1$ contains all trees with $k$ edges as subgraphs. We prove that the Erdős-Sós conjecture holds approximately, if the size of the embedded tree is linear in the size of the graph, and the maximum degree of the tree is sublinear.

math.CO↗

A version of the Loebl-Komlós-Sós conjecture for skewed trees

Loebl, Komlós, and Sós conjectured that any graph with at least half of its vertices of degree at least k contains every tree with at most k edges. We propose a version of this conjecture for skewed trees, i.e., we consider the class of trees with at most k edges such that the sizes of the colour classes of the trees have a given ratio. We show that our conjecture is asymptotically correct for dense graphs. The proof relies on the regularity method. Our result implies bounds on Ramsey number of several trees of given skew.

math.CO↗