Search

Erd\H{o}s asked whether, similarly as for graphs, the Tur\'an density of $K_4^{(3)}$ is the same as that of $K_5^{(3)-}$. This was disproved by Markstr\"om using a proof that relied on computer-aided flag algebra calculations. Here we give a flag-algebra-free proof that for all $k\geq 3$, $\pi(K_{k+1}^{(k)})<\pi(K_{k+2}^{(k)-})$. We further illustrate our methods by separating the Tur\'an densities of other hypergraphs without relying on explicit bounds. For instance, we show that $\pi(K_{\ell}^{(k)})<\pi(K_{\ell+1}^{(k)})$, for all $\ell> k\geq3$., Comment: 10 pages

We show that for every integer $k\geq3$, the set of Tur\'an densities of $k$-uniform hypergraphs has an accumulation point in $[0,1)$. In particular, $1/2$ is an accumulation point for the set of Tur\'an densities of $3$-uniform hypergraphs., Comment: 6 pages

Temporal Knowledge Graph (TKG) Forecasting aims at predicting links in Knowledge Graphs for future timesteps based on a history of Knowledge Graphs. To this day, standardized evaluation protocols and rigorous comparison across TKG models are available, but the importance of simple baselines is often neglected in the evaluation, which prevents researchers from discerning actual and fictitious progress. We propose to close this gap by designing an intuitive baseline for TKG Forecasting based on predicting recurring facts. Compared to most TKG models, it requires little hyperparameter tuning and no iterative training. Further, it can help to identify failure modes in existing approaches. The empirical findings are quite unexpected: compared to 11 methods on five datasets, our baseline ranks first or third in three of them, painting a radically different picture of the predictive quality of the state of the art., Comment: Accepted at IJCAI 2024

Advances in computer vision as well as increasingly widespread video-based behavioral monitoring have great potential for transforming how we study animal cognition and behavior. However, there is still a fairly large gap between the exciting prospects and what can actually be achieved in practice today, especially in videos from the wild. With this perspective paper, we want to contribute towards closing this gap, by guiding behavioral scientists in what can be expected from current methods and steering computer vision researchers towards problems that are relevant to advance research in animal behavior. We start with a survey of the state-of-the-art methods for computer vision problems that are directly relevant to the video-based study of animal behavior, including object detection, multi-individual tracking, individual identification, and (inter)action recognition. We then review methods for effort-efficient learning, which is one of the biggest challenges from a practical perspective. Finally, we close with an outlook into the future of the emerging field of computer vision for animal behavior, where we argue that the field should develop approaches to unify detection, tracking, identification and (inter)action recognition in a single, video-based framework.

A simplicial complex $H$ consists of a pair of sets $(V,E)$ where $V$ is a set of vertices and $E\subseteq\mathscr{P}(V)$ is a collection of subsets of $V$ closed under taking subsets. Given a simplicial complex $F$ and $n\in \mathbb N$, the extremal number $\text{ex}(n,F)$ is the maximum number of edges that a simplicial complex on $n$ vertices can have without containing a copy of $F$. We initiate the systematic study of extremal numbers in this context by asymptotically determining the extremal numbers of several natural simplicial complexes. In particular, we asymptotically determine the extremal number of a simplicial complex for which the extremal example has more than one incomplete layer., Comment: 22 pages

Let $k\geq 3$. Given a $k$-uniform hypergraph $H$, the minimum codegree $\delta(H)$ is the largest $d\in\mathbb{N}$ such that every $(k-1)$-set of $V(H)$ is contained in at least $d$ edges. Given a $k$-uniform hypergraph $F$, the codegree Tur\'an density $\gamma(F)$ of $F$ is the smallest $\gamma \in [0,1]$ such that every $k$-uniform hypergraph on $n$ vertices with $\delta(H)\geq (\gamma + o(1))n$ contains a copy of $F$. Similarly as other variants of the hypergraph Tur\'an problem, determining the codegree Tur\'an density of a hypergraph is in general notoriously difficult and only few results are known. In this work, we show that for every $\varepsilon>0$, there is a $k$-uniform hypergraph $F$ with $0<\gamma(F)<\varepsilon$. This is in contrast to the classical Tur\'an density, which cannot take any value in the interval $(0,k!/k^k)$ due to a fundamental result by Erd\H{o}s., Comment: 12 pages

Here we consider the hypergraph Tur\'an problem in uniformly dense hypergraphs as was suggested by Erd\H{o}s and S\'os. Given a $3$-graph $F$, the uniform Tur\'an density $\pi_u(F)$ of $F$ is defined as the supremum over all $d\in[0,1]$ for which there is an $F$-free uniformly $d$-dense $3$-graph, where uniformly $d$-dense means that every linearly sized subhypergraph has density at least $d$. Recently, Glebov, Kr\'al', and Volec and, independently, Reiher, R\"odl, and Schacht proved that $\pi_u(K_4^{(3)-})=\frac{1}{4}$, solving a conjecture by Erd\H{o}s and S\'os. Reiher, R\"odl, and Schacht also characterised all $3$-graphs with $\pi_u(F)=0$. Very recently, Buci\'c, Cooper, Kr\'al', Mohr, and Correira determined the uniform Tur\'an density of $3$-uniform tight cycles (of length at least $5$), answering a question by Reiher. These hypergraphs and their blow-ups are the only hypergraphs for which the uniform Tur\'an density is known. In this work, we determine the uniform Tur\'an density of the $3$-graph on five vertices that is obtained from $K_4^{(3)-}$ by adding an additional vertex whose link forms a matching on the vertices of $K_4^{(3)-}$. Further, we point to two natural intermediate problems on the way to determining $\pi_u(K_4^{(3)})$ and solve the first of these., Comment: 14 pages, 5 figures

Given $\alpha>0$ and an integer $\ell\geq5$, we prove that every sufficiently large $3$-uniform hypergraph $H$ on $n$ vertices in which every two vertices are contained in at least $\alpha n$ edges contains a copy of $C_\ell^{-}$, a tight cycle on $\ell$ vertices minus one edge. This improves a previous result by Balogh, Clemen, and Lidick\'y., Comment: 6 pages, 1 figure

Katona's intersection theorem states that every intersecting family $\mathcal F\subseteq[n]^{(k)}$ satisfies $\vert\partial\mathcal F\vert\geq\vert\mathcal F\vert$, where $\partial\mathcal F=\{F\setminus x:x\in F\in\mathcal F\}$ is the shadow of $\mathcal F$. Frankl conjectured that for $n>2k$ and every intersecting family $\mathcal F\subseteq [n]^{(k)}$, there is some $i\in[n]$ such that $\vert \partial \mathcal F(i)\vert\geq \vert\mathcal F(i)\vert$, where $\mathcal F(i)=\{F\setminus i:i\in F\in\mathcal F\}$ is the link of $\mathcal F$ at $i$. Here, we prove this conjecture in a very strong form for $n> \binom{k+1}{2}$. In particular, our result implies that for any $j\in[k]$, there is a $j$-set $\{a_1,\dots,a_j\}\in[n]^{(j)}$ such that $\vert \partial \mathcal F(a_1,\dots,a_j)\vert\geq \vert\mathcal F(a_1,\dots,a_j)\vert$. A similar statement is also obtained for cross-intersecting families., Comment: 6 pages

Subjective self-reports in VR user studies is a burdening and often tedious task for the participants. To minimize the disruption with the ongoing experience VR research has started to administer the surveying directly inside the virtual environments. However, due to the tedious nature of text-entry in VR, most VR surveying tools focus on closed questions with predetermined responses, while open questions with free-text responses remain unexplored. This neglects a crucial part of UX research. To provide guidance on suitable self-reporting methods for open questions in VR user studies, this position paper presents a comparative study with three text-entry methods in VR and outlines future directions towards low-burden qualitative responding.

The contribution of this work is twofold: (1) We introduce a collection of ensemble methods for time series forecasting to combine predictions from base models. We demonstrate insights on the power of ensemble learning for forecasting, showing experiment results on about 16000 openly available datasets, from M4, M5, M3 competitions, as well as FRED (Federal Reserve Economic Data) datasets. Whereas experiments show that ensembles provide a benefit on forecasting results, there is no clear winning ensemble strategy (plus hyperparameter configuration). Thus, in addition, (2), we propose a meta-learning step to choose, for each dataset, the most appropriate ensemble method and their hyperparameter configuration to run based on dataset meta-features., Comment: Accepted at IJCNN 2021

A famous result by R\"odl, Ruci\'nski, and Szemer\'edi guarantees a (tight) Hamilton cycle in $k$-uniform hypergraphs $H$ on $n$ vertices with minimum $(k-1)$-degree $\delta_{k-1}(H)\geq (1/2+o(1))n$, thereby extending Dirac's result from graphs to hypergraphs. For graphs, much more is known; each graph on $n$ vertices with $\delta(G)\geq (1/2+o(1))n$ contains $(1-o(1))r$ edge-disjoint Hamilton cycles where $r$ is the largest integer such that $G$ contains a spanning $2r$-regular subgraph, which is clearly asymptotically optimal. This was proved by Ferber, Krivelevich, and Sudakov answering a question raised by K\"uhn, Lapinskas, and Osthus. We extend this result to hypergraphs; every $k$-uniform hypergraph $H$ on $n$ vertices with $\delta_{k-1}(H)\geq (1/2+o(1))n$ contains $(1-o(1))r$ edge-disjoint (tight) Hamilton cycles where $r$ is the largest integer such that $H$ contains a spanning subgraph with each vertex belonging to $kr$ edges. In particular, this yields an asymptotic solution to a question of Glock, K\"uhn, and Osthus. In fact, our main result applies to approximately vertex-regular $k$-uniform hypergraphs with a weak quasirandom property and provides approximate decompositions into cycle factors without too short cycles., Comment: 27 pages

We study a Combinatorial Multi-Bandit Problem motivated by applications in energy systems management. Given multiple probabilistic multi-arm bandits with unknown outcome distributions, the task is to optimize the value of a combinatorial objective function mapping the vector of individual bandit outcomes to a single scalar reward. Unlike in single-bandit problems with multi-dimensional action space, the outcomes of the individual bandits are observable in our setting and the objective function is known. Guided by the hypothesis that individual observability enables better trade-offs between exploration and exploitation, we generalize the lower regret bound for single bandits, showing that indeed for multiple bandits it admits parallelized exploration. For our energy management application we propose a range of algorithms that combine exploration principles for multi-arm bandits with mathematical programming. In an experimental study we demonstrate the effectiveness of our approach to learn action assignments for 150 bandits, each having 24 actions, within a horizon of 365 episodes.

Given integers $r\geq 2$ and $n,t\geq 1$ we call families $\mathcal{F}_1,\dots,\mathcal{F}_r\subseteq\mathscr{P}([n])$ $r$-cross $t$-intersecting if for all $F_i\in\mathcal{F}_i$, $i\in[r]$, we have $\vert\bigcap_{i\in[r]}F_i\vert\geq t$. We obtain a strong generalisation of the classic Hilton-Milner theorem on cross intersecting families. In particular, we determine the maximum of $\sum_{j\in [r]}\vert\mathcal{F}_j\vert$ for $r$-cross $t$-intersecting families in the cases when these are $k$-uniform families or arbitrary subfamilies of $\mathscr{P}([n])$. Only some special cases of these results had been proved before. We obtain the aforementioned theorems as instances of a more general result that considers measures of $r$-cross $t$-intersecting families. This also provides the maximum of $\sum_{j\in [r]}\vert\mathcal{F}_j\vert$ for families of possibly mixed uniformities $k_1,\ldots,k_r$., Comment: 13 pages

Given two non-negative integers $n$ and $s$, define $m(n,s)$ to be the maximal number such that in every hypergraph $\mathcal{H}$ on $n$ vertices and with at most $ m(n,s)$ edges there is a vertex $x$ such that $|\mathcal{H}_x|\geq | E(\mathcal{H})| -s$, where $\mathcal{H}_x=\{H\setminus\{x\}:H\in E(\mathcal{H})\}$. This problem has been posed by F\"uredi and Pach and by Frankl and Tokushige. While the first results were only for specific small values of $s$, Frankl determined $m(n,2^{d-1}-1)$ for all $d\in\mathbb{N}$ with $d\mid n$. Subsequently, the goal became to determine $m(n,2^{d-1}-c)$ for larger $c$. Frankl and Watanabe determined $m(n,2^{d-1}-c)$ for $c\in\{0,2\}$. Other general results were not known so far. Our main result sheds light on what happens further away from powers of two: We prove that $m(n,2^{d-1}-c)=\frac{n}{d}(2^d-c)$ for $d\geq 4c$ and $d\mid n$ and give an example showing that this equality does not hold for $c=d$. The other line of research on this problem is to determine $m(n,s)$ for small values of $s$. In this line, our second result determines $m(n,2^{d-1}-c)$ for $c\in\{3,4\}$. This solves more instances of the problem for small $s$ and in particular solves a conjecture by Frankl and Watanabe., Comment: 16 pages

We show that every $k$-uniform hypergraph on $n$ vertices whose minimum $(k-2)$-degree is at least $(5/9+o(1))n^2/2$ contains a Hamiltonian cycle. A construction due to Han and Zhao shows that this minimum degree condition is optimal. The same result was proved independently by Lang and Sahueza-Matamala., Comment: Dedicated to Endre Szemer\'edi on the occasion of his 80th birthday

We investigate the problem of determining how many monochromatic trees are necessary to cover the vertices of an edge-coloured random graph. More precisely, we show that for $p\gg n^{-1/6}{(\ln n)}^{1/6}$, in any $3$-edge-colouring of the random graph $G(n,p)$ we can find three monochromatic trees such that their union covers all vertices. This improves, for three colours, a result of Buci\'c, Kor\'andi and Sudakov., Comment: 16 pages, 5 figures

We show that every 4-uniform hypergraph with $n$ vertices and minimum pair degree at least $(5/9+o(1))n^2/2$ contains a tight Hamiltonian cycle. This degree condition is asymptotically optimal., Comment: Dedicated to Endre Szemer\'edi on the occasion of his $80^{\text{th}}$ birthday

Adaptive intelligence aims at empowering machine learning techniques with the additional use of domain knowledge. In this work, we present the application of adaptive intelligence to accelerate MR acquisition. Starting from undersampled k-space data, an iterative learning-based reconstruction scheme inspired by compressed sensing theory is used to reconstruct the images. We adopt deep neural networks to refine and correct prior reconstruction assumptions given the training data. The network was trained and tested on a knee MRI dataset from the 2019 fastMRI challenge organized by Facebook AI Research and NYU Langone Health. All submissions to the challenge were initially ranked based on similarity with a known groundtruth, after which the top 4 submissions were evaluated radiologically. Our method was evaluated by the fastMRI organizers on an independent challenge dataset. It ranked #1, shared #1, and #3 on respectively the 8x accelerated multi-coil, the 4x multi-coil, and the 4x single-coil track. This demonstrates the superior performance and wide applicability of the method.

Automated algorithm selection and hyperparameter tuning facilitates the application of machine learning. Traditional multi-armed bandit strategies look to the history of observed rewards to identify the most promising arms for optimizing expected total reward in the long run. When considering limited time budgets and computational resources, this backward view of rewards is inappropriate as the bandit should look into the future for anticipating the highest final reward at the end of a specified time budget. This work addresses that insight by introducing HAMLET, which extends the bandit approach with learning curve extrapolation and computation time-awareness for selecting among a set of machine learning algorithms. Results show that the HAMLET Variants 1-3 exhibit equal or better performance than other bandit-based algorithm selection strategies in experiments with recorded hyperparameter tuning traces for the majority of considered time budgets. The best performing HAMLET Variant 3 combines learning curve extrapolation with the well-known upper confidence bound exploration bonus. That variant performs better than all non-HAMLET policies with statistical significance at the 95% level for 1,485 runs., Comment: 8 pages, 8 figures; IJCNN 2020: International Joint Conference on Neural Networks

For integers $n\geq k\geq 1$, the Kneser graph $K(n,k)$ is the graph with vertex set $V=[n]^{(k)}$ and edge set $E=\{\{x,y\} \in V^{(2)}: x\cap y=\emptyset\}$. Chen proved that for $n\geq 3k$, Kneser graphs are Hamiltonian and later improved this to $n\geq 2.62k+1$. Furthermore, Chen and F\"uredi gave a short proof that if $k | n$, Kneser graphs are Hamiltonian for $n\geq 3k$. In this note, we present a short proof that does not need the divisibility condition, i.e., we give a short proof that $K(n,k)$ is Hamiltonian for $n\geq 4k$., Comment: 3 pages