Dorina Thanou
Dorina Thanou works on graph learning and signal processing. Her coauthored 2014 paper developed learned parametric dictionaries for signals on graphs, using network structure to represent observations efficiently. She passed her PhD exam at EPFL on 21 December 2015. Her 2016 coauthored paper on heat-diffusion graphs proposed learning unknown connectivity from observed signals. It treats data as combinations of local diffusion processes and uses optimization to infer a useful graph.
Timeline
-
January 2014
Learning graph-signal dictionaries
Thanou coauthored work with David Shuman and Pascal Frossard on learning parametric dictionaries for graph signals. The method uses graph structure and learned representations for sparse approximation.
-
21 December 2015
Doctoral exam at EPFL
Thanou passed her PhD exam on 21 December 2015.
-
November 2016
Learning heat-diffusion graphs
She coauthored a method for inferring the graph connectivity that explains observed signals. The method models observations as heat-diffusion processes and uses a nonconvex optimization algorithm. Experiments cover synthetic and real data.
Sources and credits
Photo: people.epfl.ch; EPFL - Dorina Thanou; creator not stated. License: Not stated by source. Source: people.epfl.ch. Cropped to a square and resized. Ask us to remove or replace this photo.