arXiv subjects

A. Knippel

Publications and source records attributed to A. Knippel.

7 recordsLinked to original sources

arXiv2023

Eigenvectors of graph Laplacians: a landscape

We review the properties of eigenvectors for the graph Laplacian matrix, aiming at predicting a specific eigenvalue/vector from the geometry of the graph. After considering classical graphs for which the spectrum is known, we focus on eigenvectors that have zero components and extend the pioneering results of Merris (1998) on graph transformations that preserve a given eigenvalue $λ$ or shift it in a simple way. These transformations enable us to obtain eigenvalues/vectors combinatorially instead of numerically; in particular we show that graphs having eigenvalues $λ= 1,2,\dots,6$ up to six vertices can be obtained from a short list of graphs. For the converse problem of a $λ$ subgraph $G$ of a $λ$ graph $G"$, we prove results and conjecture that $G$ and $G"$ are connected by two of the simple transformations described above.

math.SP↗

arXiv2020

Epidemic model on a network: analysis and applications to COVID-19

We analyze an epidemic model on a network consisting of susceptible-infected-recovered equations at the nodes coupled by diffusion using a graph Laplacian. We introduce an epidemic criterion and examine different vaccination/containment strategies: we prove that it is most effective to vaccinate a node of highest degree. The model is also useful to evaluate deconfinement scenarios and prevent a so-called second wave. The model has few parameters enabling fitting to the data and the essential ingredient of importation of infected; these features are particularly important for the current COVID-19 epidemic.

physics.soc-ph↗

arXiv2020

Spectra of chains connected to complete graphs

We characterize the spectrum of the Laplacian of graphs composed of one or two finite or infinite chains connected to a complete graph. We show the existence of localized eigenvectors of two types, eigenvectors that vanish exactly outside the complete graph and eigenvectors that decrease exponentially outside the complete graph. Our results also imply gaps between the eigenvalues corresponding to localized and extended eigenvectors.

math.SP↗

arXiv2019

Spectral solution of load flow equations

The load-flow equations are the main tool to operate and plan electrical networks. For transmission or distribution networks these equations can be simplified into a linear system involving the graph Laplacian and the power input vector. We show, using spectral graph theory, how to solve this system efficiently. This spectral approach gives a new geometric view of the network and power vector. This formulation yields a Parseval-like relation for the $L_2$ norm of the power in the lines. Using this relation as a guide, we show that a small number of eigenvector components of the power vector are enough to obtain an estimate of the solution. This would allow fast reconfiguration of networks and better planning.

physics.class-ph↗

arXiv2018

On graph Laplacian eigenvectors with components in {-1,0,1}

We characterize all graphs for which there are eigenvectors of the graph Laplacian having all their components in {-1,+1} or {-1,0,+ 1}. Graphs having eigenvectors with components in {-1,+1} are called bivalent and are shown to be the regular bipartite graphs and their extensions obtained by adding edges between vertices with the same value for the given eigenvector. Graphs with eigenvectors with components in {-1,0,+ 1} are called trivalent and are shown to be soft-regular graphs -graphs such that vertices associated with non-zero components have the same degree- and their extensions via some transformations.

math.SP↗

arXiv2018

Localized solutions of nonlinear network wave equations

We study localized solutions for the nonlinear graph wave equation on finite arbitrary networks. Assuming a large amplitude localized initial condition on one node of the graph, we approximate its evolution by the Duffing equation. The rest of the network satisfies a linear system forced by the excited node. This approximation is validated by reducing the nonlinear graph wave equation to the discrete nonlinear Schrodinger equation and by Fourier analysis. Finally, we examine numerically the condition for localization in the parameter plane, coupling versus amplitude and show that the localization amplitude depends on the maximal normal eigenfrequency.

nlin.PS↗

arXiv2018

Inverse source problem in a forced network

We address the nonlinear inverse source problem of identifying a time-dependent source occurring in one node of a network governed by a wave equation. We prove that time records of the associated state taken at a strategic set of two nodes yield uniqueness of the two unknown elements: the source position and the emitted signal. We develop a non-iterative identification method that localizes the source node by solving a set of well posed linear systems. Once the source node is localized, we identify the emitted signal using a deconvolution problem or a Fourier expansion. Numerical experiments on a $5$ node graph confirm the effectiveness of the approach.

math.OC↗