– N. Tremblay (INRIA Rennes, EPFL Lausanne) : Filterbanks for graph signal processing

Carte non disponible

Date(s) - 9 octobre 2015

Catégories Pas de Catégories

Titre : Filterbanks for graph signal processing.\n\nRésumé : I will start with an introduction to the emerging (and exciting) field of graph signal processing. I will then raise the question of filterbanks for graph signals, and identify the main difficulties when attempting to define such filterbanks by analogy to the classical 1D discrete case. I further propose a construction that takes into account a decomposition of the graph in a partition of subgraphs : local graph Fourier bases of each subgraph are used as building blocks for the filterbanks. I finally show applications through compression and denoising experiments.\nIf time allows, I will finish with a short presentation of my latest research project : accelerating the spectral clustering algorithm by graph filtering random signals.\nHope to see you there ![