Transformation de graphes dynamiques en signaux non stationnaires

Abstract : Many data associated to networks, whether physical, biological or social, can be described by graphs that can become dynamic if a time evolution is added. These graphs are difficult to study because there exist only a few tools to describe these objects. The objective here is to propose a new method to visualize synthetically their evolution over time. The originality of this work consists of study dynamic graphs using a signal theory approach, by computing spectral analysis on series representing the graphs. The method consists of transforming a graph to a collection of signals using multidimensional scaling then linking frequency patterns of these series with graph properties. The extension to dynamic graph enables us to follow the evolution of these patterns and then track the modification of the structure of the graph over time. A method to reconstruct the graph from the collection of signals is also proposed to reduce the graph by selecting the most significant links.
Complete list of metadatas

https://hal-ens-lyon.archives-ouvertes.fr/ensl-00875085
Contributor : Ronan Hamon <>
Submitted on : Monday, October 21, 2013 - 9:58:22 AM
Last modification on : Tuesday, February 26, 2019 - 4:19:03 PM
Long-term archiving on : Friday, April 7, 2017 - 1:30:20 PM

File

article_gretsi2013.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : ensl-00875085, version 1

Citation

Ronan Hamon, Pierre Borgnat, Patrick Flandrin, Céline Robardet. Transformation de graphes dynamiques en signaux non stationnaires. Colloque GRETSI 2013, Sep 2013, Brest, France. pp.251. ⟨ensl-00875085⟩

Share

Metrics

Record views

403

Files downloads

179