Showing 1–8 of 8 results for author: de Kergorlay, H

Connectivity of Random Geometric Hypergraphs

Authors: Henry-Louis de Kergorlay, Desmond J. Higham

Abstract: We consider a random geometric hypergraph model based on an underlying bipartite graph. Nodes and hyperedges are sampled uniformly in a domain, and a node is assigned to those hyperedges that lie with a certain radius. From a modelling perspective, we explain how the model captures higher order connections that arise in real data sets. Our main contribution is to study the connectivity properties… ▽ More We consider a random geometric hypergraph model based on an underlying bipartite graph. Nodes and hyperedges are sampled uniformly in a domain, and a node is assigned to those hyperedges that lie with a certain radius. From a modelling perspective, we explain how the model captures higher order connections that arise in real data sets. Our main contribution is to study the connectivity properties of the model. In an asymptotic limit where the number of nodes and hyperedges grow in tandem we give a condition on the radius that guarantees connectivity. △ Less

Submitted 17 September, 2023; originally announced September 2023.

MSC Class: 05C80 ACM Class: G.2.2

Estimating Network Dimension When the Spectrum Struggles

Authors: Peter Grindrod, Desmond John Higham, Henry-Louis de Kergorlay

Abstract: What is the dimension of a network? Here, we view it as the smallest dimension of Euclidean space into which nodes can be embedded so that pairwise distances accurately reflect the connectivity structure. We show that a recently proposed and extremely efficient algorithm for data clouds, based on computing first and second nearest neighbour distances, can be used as the basis of an approach for es… ▽ More What is the dimension of a network? Here, we view it as the smallest dimension of Euclidean space into which nodes can be embedded so that pairwise distances accurately reflect the connectivity structure. We show that a recently proposed and extremely efficient algorithm for data clouds, based on computing first and second nearest neighbour distances, can be used as the basis of an approach for estimating the dimension of a network with weighted edges. We also show how the algorithm can be extended to unweighted networks when combined with spectral embedding. We illustrate the advantages of this technique over the widely-used approach of characterising dimension by visually searching for a suitable gap in the spectrum of the Laplacian. △ Less

Submitted 25 June, 2023; originally announced June 2023.

MSC Class: 05C20; 05C80; 05C85; 05C90; 05C82 ACM Class: G.2.2

Mean Field Analysis of Hypergraph Contagion Model

Authors: Desmond J. Higham, Henry-Louis de Kergorlay

Abstract: We typically interact in groups, not just in pairs. For this reason, it has recently been proposed that the spread of information, opinion or disease should be modelled over a hypergraph rather than a standard graph. The use of hyperedges naturally allows for a nonlinear rate of transmission, in terms of both the group size and the number of infected group members, as is the case, for example, whe… ▽ More We typically interact in groups, not just in pairs. For this reason, it has recently been proposed that the spread of information, opinion or disease should be modelled over a hypergraph rather than a standard graph. The use of hyperedges naturally allows for a nonlinear rate of transmission, in terms of both the group size and the number of infected group members, as is the case, for example, when social distancing is encouraged. We consider a general class of individual-level, stochastic, susceptible-infected-susceptible models on a hypergraph, and focus on a mean field approximation proposed in [Arruda et al., Phys. Rev. Res., 2020]. We derive spectral conditions under which the mean field model predicts local or global stability of the infection-free state. We also compare these results with (a) a new condition that we derive for decay to zero in mean for the exact process, (b) conditions for a different mean field approximation in [Higham and de Kergorlay, Proc. Roy. Soc. A, 2021], and (c) numerical simulations of the microscale model. △ Less

Submitted 11 August, 2021; originally announced August 2021.

MSC Class: 92C60; 37N25; 05C65

Which Sampling Densities are Suitable for Spectral Clustering on Unbounded Domains?

Authors: Henry-Louis de Kergorlay

Abstract: We consider a random geometric graph with vertices sampled from a probability measure supported on $\mathbb R^d$, and study its connectivity. We show the graph is typically disconnected, unless the sampling density has superexponential decay. In the later setting, we identify an asymptotic threshold value for the radius parameter of the graph such that, for radius values beyond the threshold, some… ▽ More We consider a random geometric graph with vertices sampled from a probability measure supported on $\mathbb R^d$, and study its connectivity. We show the graph is typically disconnected, unless the sampling density has superexponential decay. In the later setting, we identify an asymptotic threshold value for the radius parameter of the graph such that, for radius values beyond the threshold, some concentration properties hold for the sampled points of the graph, while the graph is disconnected for radius values below the same threshold. Properties of point processes are well-known to be closely related to the analysis of geometric learning problems, such as spectral clustering. This work can be seen as a first step towards understanding the consistency of spectral clustering when the probability measure has unbounded support. In particular, we narrow down the setting under which spectral clustering algorithms on $\mathbb R^d$ may be expected to achieve consistency, to a sufficiently fast decay of the sampling density (superexponential) and a sufficiently slowly decaying radius parameter value as a function of $n$, the number of sampled points. △ Less

Submitted 5 April, 2021; originally announced April 2021.

Epidemics on Hypergraphs: Spectral Thresholds for Extinction

Authors: Desmond John Higham, Henry-Louis de Kergorlay

Abstract: Epidemic spreading is well understood when a disease propagates around a contact graph. In a stochastic susceptible-infected-susceptible setting, spectral conditions characterise whether the disease vanishes. However, modelling human interactions using a graph is a simplification which only considers pairwise relationships. This does not fully represent the more realistic case where people mee… ▽ More Epidemic spreading is well understood when a disease propagates around a contact graph. In a stochastic susceptible-infected-susceptible setting, spectral conditions characterise whether the disease vanishes. However, modelling human interactions using a graph is a simplification which only considers pairwise relationships. This does not fully represent the more realistic case where people meet in groups. Hyperedges can be used to record such group interactions, yielding more faithful and flexible models, allowing for the rate of infection of a node to vary as a nonlinear function of the number of infectious neighbors. We discuss different types of contagion models in this hypergraph setting, and derive spectral conditions that characterize whether the disease vanishes. We study both the exact individual-level stochastic model and a deterministic mean field ODE approximation. Numerical simulations are provided to illustrate the analysis. We also interpret our results and show how the hypergraph model allows us to distinguish between contributions to infectiousness that (a) are inherent in the nature of the pathogen and (b) arise from behavioural choices (such as social distancing, increased hygiene and use of masks). This raises the possibility of more accurately quantifying the effect of interventions that are designed to contain the spread of a virus. △ Less

Submitted 12 March, 2021; originally announced March 2021.

MSC Class: 92C60; 37N25; 05C65 ACM Class: J.4; J.3

Consistency of Anchor-based Spectral Clustering

Authors: Henry-Louis de Kergorlay, Desmond John Higham

Abstract: Anchor-based techniques reduce the computational complexity of spectral clustering algorithms. Although empirical tests have shown promising results, there is currently a lack of theoretical support for the anchoring approach. We define a specific anchor-based algorithm and show that it is amenable to rigorous analysis, as well as being effective in practice. We establish the theoretical consisten… ▽ More Anchor-based techniques reduce the computational complexity of spectral clustering algorithms. Although empirical tests have shown promising results, there is currently a lack of theoretical support for the anchoring approach. We define a specific anchor-based algorithm and show that it is amenable to rigorous analysis, as well as being effective in practice. We establish the theoretical consistency of the method in an asymptotic setting where data is sampled from an underlying continuous probability distribution. In particular, we provide sharp asymptotic conditions for the algorithm parameters which ensure that the anchor-based method can recover with high probability disjoint clusters that are mutually separated by a positive distance. We illustrate the performance of the algorithm on synthetic data and explain how the theoretical convergence analysis can be used to inform the practical choice of parameter scalings. We also test the accuracy and efficiency of the algorithm on two large scale real data sets. We find that the algorithm offers clear advantages over standard spectral clustering. We also find that it is competitive with the state-of-the-art LSC method of Chen and Cai (Twenty-Fifth AAAI Conference on Artificial Intelligence, 2011), while having the added benefit of a consistency guarantee. △ Less

Submitted 27 June, 2020; v1 submitted 24 June, 2020; originally announced June 2020.

Random Čech complexes on $\mathbb{R}^d$: decrackling the noise with local scalings

Authors: Henry-Louis de Kergorlay

Abstract: We investigate the homology of an unbounded noisy sample on $\mathbb{R}^d$, under various assumptions on the sampling density. This investigation is based on previous results by Adler, Bobrowski, and Weinberger (\cite{crackle}), and Owada and Adler (\cite{topoCrackle}). There, it was found that unbounded noise generally introduces non-vanishing homology, a phenomenon called \textit{topological cra… ▽ More We investigate the homology of an unbounded noisy sample on $\mathbb{R}^d$, under various assumptions on the sampling density. This investigation is based on previous results by Adler, Bobrowski, and Weinberger (\cite{crackle}), and Owada and Adler (\cite{topoCrackle}). There, it was found that unbounded noise generally introduces non-vanishing homology, a phenomenon called \textit{topological crackle}, unless the density has superexponential decay on $\mathbb{R}^d$. We show how some well-chosen \textit{non-trivial} variable bandwidth constructions can extend the class of densities where crackle doesn't occur to any light tail density with mild assumptions, what we call \textit{decrackling the noise}. △ Less

Submitted 22 December, 2019; originally announced December 2019.

Random Čech Complexes on Manifolds with Boundary

Authors: Henry-Louis de Kergorlay, Ulrike Tillmann, Oliver Vipond

Abstract: Let $M$ be a compact, unit volume, Riemannian manifold with boundary. In this paper we study the homology of a random Čech-complex generated by a homogeneous Poisson process in $M$. Our main results are two asymptotic threshold formulas, an upper threshold above which the Čech complex recovers the $k$-th homology of $M$ with high probability, and a lower threshold below which it almost certainly d… ▽ More Let $M$ be a compact, unit volume, Riemannian manifold with boundary. In this paper we study the homology of a random Čech-complex generated by a homogeneous Poisson process in $M$. Our main results are two asymptotic threshold formulas, an upper threshold above which the Čech complex recovers the $k$-th homology of $M$ with high probability, and a lower threshold below which it almost certainly does not. These thresholds are close together in the sense that they have the same leading term. Here $k$ is positive and strictly less than the dimension $d$ of the manifold. This extends work of Bobrowski and Weinberger in [BW17] and Bobrowski and Oliveira [BO19] who establish similar formulas when $M$ is a torus and, more generally, is closed and has no boundary. We note that the cases with and without boundary lead to different answers: The corresponding common leading terms for the upper and lower thresholds differ being $\log (n) $ when $M$ is closed and $(2-2/d)\log (n)$ when $M$ has boundary; here $n$ is the expected number of sample points. Our analysis identifies a special type of homological cycle, which we call a $Θ$-like-cycle, which occur close to the boundary and establish that the first order term of the lower threshold is $(2-2/d)\log (n)$. △ Less

Submitted 18 June, 2019; originally announced June 2019.

Comments: 42 pages, 8 figures

MSC Class: 60B99; 60D05; 05E45; 55U10; 53B21