Blind Graph Topology Change Detection

Journal Article (2018)
Author(s)

E. Isufi (TU Delft - Signal Processing Systems)

Ashvant S.U. Mahabir (Student TU Delft)

G Leus (TU Delft - Signal Processing Systems)

Research Group
Signal Processing Systems
Copyright
© 2018 E. Isufi, Ashvant S.U. Mahabir, G.J.T. Leus
DOI related publication
https://doi.org/10.1109/LSP.2018.2819127
More Info
expand_more
Publication Year
2018
Language
English
Copyright
© 2018 E. Isufi, Ashvant S.U. Mahabir, G.J.T. Leus
Research Group
Signal Processing Systems
Issue number
5
Volume number
25
Pages (from-to)
655-659
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

This letter investigates methods to detect graph topological changes without making any assumption on the nature of the change itself. To accomplish this, we merge recently developed tools in graph signal processing with matched subspace detection theory and propose two blind topology change detectors. The first detector exploits the prior information that the observed signal is sparse w.r.t. the graph Fourier transform of the nominal graph, while the second makes use of the smoothness prior w.r.t. the nominal graph to detect topological changes. Both detectors are compared with their respective nonblind counterparts in a synthetic scenario that mimics brain networks. The absence of information about the alternative graph, in some cases, might heavily influence the blind detector's performance. However, in cases where the observed signal deviates slightly from the nonblind model, the information about the alternative graph turns out to be not useful.

Files

Blind_Graph_Topology_Change_De... (pdf)
(pdf | 0.571 Mb)
- Embargo expired in 30-11-2018
License info not available