A series of thematic workshops in memory of Alex B. Gershman
STATOS Thematic Workshop 2018
Graph Signal Processing
Friday, September 7, 2018
2pm – 6:30pm
(right after EUSIPCO 2018)
Centro Congressi di Confindustria
– Auditorium della Tecnica
Via dell’Astronomia 30, 00146, Rome, Italy.
(venue oft he EUSIPCO)
Call for participation:
Maria Sabrina Greco, University of Pisa
Geert Leus, Delft
Prof. Sergio Barbarossa
Time: 2:00 pm – 3:00 pm
Title: „Living in a World of Multiway and Uncertain Relations: Signal Processing over Discrete Sets“
Sergio Barbarossa (S’84–M’88–F’12) received the M.S. degree and Ph.D. degree in EE from the Sapienza University of Rome, Rome, Italy, where he is currently a Full Professor. He has held visiting positions at the Environmental Research Institute of Michigan (’88), University of Virginia (’95, ’97), and University of Minnesota (’99). He has been the scientific coordinator of several EU projects on wireless sensor networks, small cell networks, and distributed mobile cloud computing. He is currently the Technical Manager of the H2020 Europe/Japan project 5G-Mi-Edge. His current research interests include the area of graph signal processing, distributed optimization, millimeter wave communications, mobile edge computing, and machine learning. He is an EURASIP Fellow and he has been an IEEE Distinguished Lecturer. He received the 2000 and 2014 IEEE Best Paper Awards from the IEEE Signal Processing Society and the 2010 Technical Achievements Award from the EURASIP
Gonzalo Mateos Buckstein
Time: 3:00 pm – 4:00 pm
Title: “Network topology inference from spectral templates”
Advancing a holistic theory of networks necessitates fundamental breakthroughs in modeling, identiﬁcation, and controllability of distributed network processes – often conceptualized as signals deﬁned on the vertices of a graph. Under the assumption that the signal properties are related to the topology of the graph where they are supported, the goal of graph signal processing (GSP) is to develop algorithms that fruitfully leverage this relational structure, and can make inferences about these relationships when they are only partially observed.
After presenting the fundamentals of GSP, we leverage these ideas to address the problem of network topology inference from graph signal observations. It is assumed that the unknown graph encodes direct relationships between signal elements, which we aim to recover from observable indirect relationships generated by a diffusion process on the graph. The innovative approach is to consider the Graph Fourier Transform of the acquired signals associated with an arbitrary graph and, among all the feasible networks, search for one that endows the resulting transforms with target spectral properties and the sought graph with appealing physical characteristics such as sparsity. Leveraging results from GSP and sparse recovery, efficient topology inference algorithms with theoretical guarantees are put forth. Numerical tests corroborate de effectiveness of the proposed algorithms when used to recover social and structural brain networks from synthetically-generated signals, as well as to identify the structural properties of proteins.
Gonzalo Mateos earned the B.Sc. degree from Universidad de la Republica, Uruguay, in 2005, and the M.Sc. and Ph.D. degrees from the University of Minnesota, Twin Cities, in 2009 and 2011, all in electrical engineering. He joined the University of Rochester, Rochester, NY, in 2014, where he is currently an Assistant Professor with the Department of Electrical and Computer Engineering, as well as a member of the Goergen Institute for Data Science. During the 2013 academic year, he was a visiting scholar with the Computer Science Department at Carnegie Mellon University. From 2004 to 2006, he worked as a Systems Engineer at Asea Brown Boveri (ABB), Uruguay. His research interests lie in the areas of statistical learning from Big Data, network science, decentralized optimization, and graph signal processing, with applications in dynamic network health monitoring, social, power grid, and Big Data analytics. He currently serves as Associate Editor for the IEEE Transactions on Signal Processing, the EURASIP Journal on Advances in Signal Processing, and is a member of the IEEE SigPort Editorial Board. Dr. Mateos received the NSF CAREER Award in 2018, the 2017 IEEE Signal Processing Society Young Author Best Paper Award (as senior co-author), and the Best Paper Awards at ICASSP 2018, SSP Workshop 2016, and SPAWC 2012. His doctoral work has been recognized with the 2013 University of Minnesota's Best Dissertation Award (Honorable Mention) across all Physical Sciences and Engineering areas.
Prof. Dimitri Van De Ville
Time: 4:30 pm – 5:30 pm
Title: “Graph Signal Processing Opens New Perspectives for Human Brain Imaging”
State-of-the-art magnetic resonance imaging (MRI) provides unprecedented opportunities to study brain structure (anatomy) and function (physiology). Based on such data, graph representations can be built where nodes are associated to brain regions and edge weights to strengths of structural or functional connections. In particular, structural graphs capture major neural pathways in white matter, while functional graphs map out statistical interdependencies between pairs of regional activity traces. Network analysis of these graphs has revealed emergent system-level properties of brain structure or function, such as efficiency of communication and modular organization.
In this talk, graph signal processing (GSP) will be presented as a novel framework to integrate brain structure, contained in the structural graph, with brain function, characterized by activity traces that can be considered as time-dependent graph signals. Such a perspective allows to define novel meaningful graph-filtering operations of brain activity that take into account the anatomical backbone. For instance, we will show how activity can be analyzed in terms of being aligned versus liberal with respect to brain structure, or how additional prior information about cognitive systems can be incorporated. The well-known Fourier phase randomization method to generate surrogate data can also be adapted to this new setting. Finally, recent work will highlight how the spatial resolution of this type of analyses can be increased to the voxel level, representing a few ten thousands of nodes.
Dimitri Van De Ville received his M.S. and Ph.D. degrees in Computer Science from Ghent University, Belgium in 1998 and 2002, respectively. From 2002 to 2005, he was a post-doctoral fellow at the Biomedical Imaging Group of Prof. Michael Unser at the Ecole Polytechnique Fédérale de Lausanne (EPFL), Switzerland. In 2005, he became responsible for the Signal Processing Unit at the University of Geneva (UniGE) and Geneva University Hospital (HUG) as part of the Centre d’Imagerie BioMédicale (CIBM), a large imaging initiative of the Lemanic academic institutions. In 2009, he was awarded an SNSF professorship and he started a joint tenure-track professorship at the UniGE (Department of Radiology and Medical Informatics, Faculty of Medicine) and the EPFL (Institute of Bioengineering, School of Engineering). Since 2015 he’s Associate Professor of Bioengineering at both institutions and his lab is located at the newly established Campus Biotech in Geneva.
He has published more than 100 journal papers on signal and image processing, including on wavelet theory and network science, and their application to the biomedical field, in particular functional brain imaging. Recent work on dynamic functional connectivity included evidence that resting-state functional networks can be disentangled in terms of their temporal overlap, which showed a more complete picture of dynamic organization of brain function and opens avenues for more sensitive biomarkers; e.g., in early diagnosis of patients with mild cognitive impairment. He is also interested in using real-time fMRI for neurofeedback applications.
Prof. Vincent Gripon
Time: 5:30 pm – 6:30 pm
Title: “Matching Convolutional Neural Networks with Graph Signals”
Multi-layer perceptrons are on the paper a generalization of Convolutional Neural Networks (CNNs). But on regular signals such as images or sounds, the latter typically perform much better than the former. This is because CNNs take into account priors about the signals: stationarity, compositionality and spatial coherence. An increasing number of datasets contain signals for which those priors are not explicitly well defined. Examples of such signals are: sensor networks, point clouds, neuroimagery, social trends, citation networks. Proposing methods that aim at leveraging the underlying structure in these settings has thus become a very active field of research.
In this talk, we will overview the various methods proposed in the literature. Some methods make use of the graph signal processing mathematical framework and define operators in the spectral domain of the graphs that represent the underlying structure of the signals. Some other methods directly tackle the problem in the vertex domain and aim at extending convolution to signals on graphs. We will mainly consider three problems: classification of signals on graphs, classification of graphs and semi-supervised classification of vertices in a graph.
Vincent Gripon (S’10-M’12) received the MS degree from Ecole Normale Superieure of Cachan and the PhD degree from Telecom Bretagne. He is a permanent researcher with Telecom Bretagne (Institut Mines-Telecom), Brest, France. His research interests lie at the intersection of information theory, computer science, and neural networks. He is the co-creator and organizer of an online programming contest named TaupIC which targets French top undergraduate students. He is a member of the IEEE.
Co-organized by EURASIP’s Special Area Team on Signal Processing for Multisensor Systems (SPMus) and the Association for the Promotion of Science and Education, a German nonprofit organization supporting children’s education.