The workshop will consist in poster sessions, presentations and invited talks.

### Information for poster sessions

- Poster boards will be 180cm (vertical) x 120cm (horizontal) and are mostly adapted to A0 format. Adhesive tape will be available on site.
- The poster area will be located close to the conference room. Poster sessions will last 1h30mn each, but it is suggested to display posters for the entire day.

### Information for oral sessions

- Oral presentations are allocated slots of 20 minutes in total. Typically, a talk should last between 15 to 17 minutes, in order to save time for questions. Please make sure everything is ready for your presentation before your session starts.

## Schedule

### Wed. June 6th

**12:30 – 13:30**

Welcome coffee

**13:30 – 15:00**

Oral session: GSP Theory 1

**15:00 – 15:30**

Coffee break

**15:30 – 16:30**

Invited talk: E. Kolaczyk

Slides

Video

**16:30 – 18:00**

Poster session: GSP Theory 2

### Thu. June 7th

**9:00 – 10:00**

Invited talk: M. Bronstein

Slides

Video

**10:00 – 10:30**

Coffee break

**10:30 – 11:30**

Oral session: GSP for Machine Learning 1

**11:30 – 12:30**

Oral session: GSP Theory 3

**12:30 – 14:00**

Lunch break

**14:00 – 15:00**

Invited talk: D. Van de Ville

Slides

Video

**15:00 – 15:30**

Coffee break

**15:30 – 16:15**

Oral session: GSP Applications 1

**16:15 – 17:45**

Poster session: GSP for Machine Learning 2

**18:00 – 20:00**

Social event at “Le Débarcadère”

### Fri. June 8th

**9:00 – 10:00**

Invited talk: G. Cheung

Slides

Video

**10:00 – 10:30**

Coffee break

**10:30 – 12:00**

Poster session: GSP Applications 2

**12:30 – 14:00**

Lunch break

**14:00 – 15:00**

Invited talk: G. Puy

Slides

Video

**15:00 – 15:30**

Coffee break

**15:30 – 17:00**

Oral session: GSP for Machine Learning 3

## Invited Speakers

- Eric D. Kolaczyk, Boston University
Abstract**Why Aren’t Network Statistics Accompanied By Uncertainty Statements?**

Over 500K articles have been published since 1999 with the word “network” in the title. And the vast majority of these report network summary statistics of one type or another. However, these numbers are rarely accompanied by any quantification of uncertainty. Yet any error inherent in the measurements underlying the construction of the network, or in the network construction procedure itself, necessarily must propagate to any summary statistics reported. Perhaps surprisingly, there is little in the way of statistical methodology for this problem. I summarize results from our recent work, for the case of estimating the density of low-order subgraphs. Under a simple model of network error, we show that consistent estimation of such densities is impossible when the rates of error are unknown and only a single network is observed. We then develop method-of-moment estimators of subgraph density and error rates for the case where a minimal number of network replicates are available. These estimators are shown to be asymptotically normal as the number of vertices increases to infinity. We also provide confidence intervals for quantifying the uncertainty in these estimates based on the asymptotic normality. We illustrate the use of our estimators in the context of gene coexpression networks. This is joint work with Qiwei Yao and Jinyuan Chang.Video

- Dimitri van de Ville, EPFL/University of Geneva
**A Graph Signal Processing Perspective on Functional Brain Imaging**

AbstractState-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 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 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.Video

- Gene Cheung, NII, Tokyo

**Recent Advances in Graph-Spectral Image Processing**

AbstractWhile much effort in graph signal processing (GSP) has focused on signals that live naturally on irregular data kernels described by graphs, GSP techniques can also be effectively applied to imaging data that reside on 2D grids. We focus the talk on recent advances in three application areas of GSP for imaging: i) compression, ii) restoration, iii) filtering. For compression, we overview the use of graph lifting transform (GLT) and signed graph Fourier transform (SGFT) for compression of 3D imaging data like light field and 3D point cloud. For restoration, we discuss the use of (reweighted) graph total variation (GTV) for denoising of 3D point cloud and blind deblurring of natural images. For filtering, we show how L1-spectral filters—efficiently implemented via the Lanczos algorithm—can flexibly be used for a range of edge-adaptive image filtering problems. Finally, we discuss future direction of unrolling graph-based iterative algorithms into neural net layers for efficient implementation and robustness guarantees for inverse imaging problems.Video - Gilles Puy, Technicolor, Rennes

**Structured sampling and fast reconstruction of smooth graph signals**

AbstractThis work concerns sampling of smooth signals on arbitrary graphs. We first study a structured sampling strategy for such smooth graph signals that consists of a random selection of few pre-defined groups of nodes. The number of groups to sample to stably embed the set ofVideo*k*-bandlimited signals is driven by a quantity called the group graph cumulative coherence. For some optimised sampling distributions, we show that sampling O(*k*log(*k*)) groups is always sufficient to stably embed the set of*k*-bandlimited signals but that this number can be smaller — down to O(log(*k*)) — depending on the structure of the groups of nodes. Fast methods to approximate these sampling distributions are detailed. Second, we consider*k*-bandlimited signals that are nearly piecewise constant over pre-defined groups of nodes. We show that it is possible to speed up the reconstruction of such signals by reducing drastically the dimension of the vectors to reconstruct. When combined with the proposed structured sampling procedure, we prove that the method provides stable and accurate reconstruction of the original signal. Finally, we present numerical experiments that illustrate our theoretical results and, as an example, show how to combine these methods for interactive object segmentation in an image using superpixels. - Michael Bronstein, Università della Svizzera Italiana, Lugano

**Geometric Deep Learning on Graphs and Manifolds: Going Beyond Euclidean Data**

AbstractIn the past decade, deep learning methods have achieved unprecedented performance on a broad range of problems in various fields from computer vision to speech recognition. So far research has mainly focused on developing deep learning methods for Euclidean-structured data. However, many important applications have to deal with non-Euclidean structured data, such as graphs and manifolds. Such data are becoming increasingly important in computer graphics and 3D vision, sensor networks, drug design, biomedicine, high energy physics, recommendation systems, and web applications. The adoption of deep learning in these fields has been lagging behind until recently, primarily since the non-Euclidean nature of objects dealt with makes the very definition of basic operations used in deep networks rather elusive. In this talk, I will introduce the emerging field of geometric deep learning on graphs and manifolds, overview existing solutions and applications, and outline the key difficulties and future research directions.Video

### Oral session – GSP Theory 1 (chair: J. Moura)

List of papers- Edge-Variant Graph Filters –
*G. Leus, M. Coutino and E. Isufi, TU Delft* - Design of Sampling Matrices in Graph Frequency Domain Taking a Trade-Off for Characteristics Between Vertex and Frequency Domains –
*Y. Shimizu and Y. Tanaka, Tokyo University of Agriculture and Technology; S. Ono, Tokyo Institute of Technology* - A Scalable M-Channel Critically Sampled Filter Bank for Graph Signals –
*D. Shuman, Macalester College* - Sampling of graph signals via randomized local aggregations –
*D. Valsesia, G. Fracastoro and E. Magli, Politecnico di Torino* - Brain Signal Markers of Behavior Are Uncovered with Graph Signal Processing Tools –
*A. Ribeiro, University of Pennsylvania*

### Poster session – GSP Theory 2

List of papers- A-Optimal Sampling and Robust Reconstruction for Graph Signals via Truncated Neumann Series –
*F. Wang, Xidian University; G. Cheung, National Institute of Informatics* - Graph-Projected Signal Processing –
*V. Gripon, IMT Atlantique* - Observing Bandlimited Graph Processes –
*E. Isufi and G. Leus, TU Delft; P. Banelli, University of Perugia; P. Di Lorenzo, Sapienza University of Rome* - Towards a spectral characterization of graph coarsening –
*A. Loukas and P. Vandergheynst, École polytechnique Fédérale de Lausanne* - Fast Decentralized Projections via Graph Filters –
*D. Romero, S. Mollaebrahim, T. Weerasinghe, C. Asensio-Marco and B. Beferull-Lozano, University of Agder* - Almost Tight Spectral Graph Wavelets with Polynomial Filters –
*D. Tay, Deakin University; Y. Tanaka and A. Sakiyama, Tokyo University of Agriculture and Technology* - Optimization over directed graphs: Linear algorithm with geometric convergence –
*R. Xin and U. Khan, Tufts University* - Multidimensional Analytic Signal with Application on Graphs –
*M. Tsitsvero, P. Borgnat and P. Goncalves, ENS de Lyon* - Graph Syndromes based Sampling of Band-limited Graph Signals –
*A. K. Achanna and G. C. Mariswamy, Tata Consultancy Services; K. Kriti, TCS Research and Innovation* - On Sparse Graph Fourier Transform –
*S. H. Safavi amd F. Torkamani-Azar, Shahid Beheshti University; M. Khatua and N.-M. Cheung, Singapore University of Technology and Design* - Frequency interpretation of random walk subspaces on directed graphs –
*H. Sevi, CEA/ENS Lyon; G. Rilling, CEA; P. Borgnat, CNRS, ENS Lyon* - A Graphical Approach to Learning Networked Parameters –
*A. Tajer, RPI*

### Oral session – GSP for Machine Learning 1 (chair: D. Shuman)

List of papers- Predicting Under and Overfitting in Deep Neural Networks using Graph Smoothness –
*C. E. Rosar Kos Lassance and V. Gripon, IMT Atlantique; A. Ortega, USC* - Aggregation Convolutional Neural Networks for Graph Signals –
*F. Gama and A. Ribeiro, University of Pennsylvania; A. Marques, King Juan Carlos University; G. Leus, TU Delft* - Advances in Deep Learning on Graphs –
*M. Defferrard, Ecole Polytechnique Fédérale de Lausanne*

### Oral session – GSP Theory 3 (chair: A. Marques)

List of papers- Diffusion LMS Strategies for Adaptive Graph Signal Processing –
*F. Hua, Laboratoire Lagrange, Université Côte d’Azur; R. Nassif and A. H. Sayed, Ecole Polytechnique Fédérale de Lausanne; C. Richard, University Nice Sophia Antipolis; H. Wang, Northwestern Polytechnical University at Xi’an* - Irregularity-Aware Graph Fourier Transforms –
*B. Girault, A. Ortega and S. Narayanan, University of Southern California* - Signal Processing over Simplicial Complexes –
*S. Barbarossa, S. Sardellitti and E. Ceci, Sapienza University of Rome*

### Oral session – GSP Applications 1 (chair: A. Ortega)

List of papers- A review on Graph Signal Processing for Neuroimaging – reinterpreting the signal processing analogy in a cognitive neuroscience perspective –
*N. Farrugia, IMT Atlantique* - Graph Signal Representation of EEG for Graph Convolutional Neural Networks –
*S. Jang and J.S. Lee, Yonsei University*

### Poster session – GSP for Machine Learning 2

List of papers- Blind Graph Topology Change Detection –
*E. Isufi and G. Leus, TU Delft* - Semi-Supervised Learning of Anomalies and Signals over Multilayer Graphs –
*V. Ioannidis and G. B. Giannakis, University of Minnesota* - Matching Convolutional Neural Networks without Priors about Data –
*C. E. Rosar Kos Lassance, J.-C. Vialatte, V. Gripon and N. Farrugia, IMT Atlantique* - An extension of convolutional neural networks to irregular domains through graph inference and translations identification – B
*. Pasdeloup, EPFL; J.-C.Vialatte and V. Gripon, IMT Atlantique* - Recommender Systems via Graph Signal Processing –
*W. Huang and A. Ribeiro, University of Pennsylvania; A. Marques, King Juan Carlos University* - Value Function Approximation in Graph-based Reinforcement Learning –
*S. Madjiheurem, University College London; L. Toni, UCL* - Inferring Mobile App Preference via Multi-View Geometric Information Fusion –
*Y. Leng, MIT Media Lab* - Community detection with the triplet loss –
*A. Galland, INRIA; M. Lelarge, INRIA-ENS* - Blind Community Detection from Low-rank Excitations of a Graph Filter –
*H. To Wai and A. Scaglione, Arizona State University; S. Segarra, A. Ozdaglar and A. Jadbabaie, MIT* - Low-Rank Decomposition for Signals on Product Graphs –
*R. Varma and J. Kovacevic, Carnegie Mellon University* - Graph learning with missing data –
*R. Liu and N.-M. Cheung, Singapore University of Technology and Design* - Characterizing Anomalous Subgraphs through Description Length –
*A. Høst-Madsen and J. Zhang, University of Hawaii* *L*-PageRank for Graph-Based Semi-Supervised Learning –^{2}*E. Bautista, ENS de Lyon*

### Poster session – GSP Applications 2

List of papers- Rate Distortion Optimized Graph Partitioning for Omnidirectional Image Coding –
*M. Rizkallah, IRISA,Univ. of Rennes 1; F. De Simone and P. Frossard, EPFL; T. Maugey and C. Guillemot, INRIA* - Reweighted Graph
*l*-Spectral Filtering –_{1}*Y. Bai and W. Gao, Peking University; G. Cheung, National Institute of Informatics; X. Liu, Harbin Institute of Technology* - Estimating the Topology of Neural Networks from Distributed Observations –
*R. Alexandru, P. Malhotra, S. Reynolds and P. L. Dragotti, Imperial College London* - p-Laplacians on Directed Graphs –
*Z. Abu-Aisheh, S. Bougleux and O. Lézoray, University of Caen* - The nonlocal p-Laplacian Variational Problem on Graphs: The Continuum Limit –
*Y. Hafiene* - Pattern extraction in multi-trial dynamical graphs of functional connectivities –
*G. Frusque, P. Borgnat and P. Goncalves, ENS de Lyon* - Estimating functional connectivity from fMRI data using a frequency-sparse multivariate autoregressive model –
*A. Bohannon, US Army Research Laboratory* - Ambiguity resolution in polarimetric multi-view stereo –
*A. K. Achanna, G. C. Mariswamy and P. Balamuralidhar, Tata Consultancy Services* - Graph signal processing for the brain: local versus global network interactions during resting state –
*M. G. Preti and D. Van De Ville, Ecole Polytechnique Fédérale de Lausanne* - Node-level Anomaly Detection in Communication Networks –
*B. Le Bars and A. Kalogeratos, CMLA – ENS Paris-Saclay* - Facial expression recognition with enhanced feature extraction using graph Fourier transform –
*M. K. Hemant and S. Kamalesh, MNIT Jaipur; S. D. Joshi, Indian Institute of Technology Delhi* - Extrapolating white matter structures via structurally-informed graph diffusion –
*A. Tarun and D. Van De Ville, École Polytechnique Fédérale de Lausanne* - Classifying non-stationary time-varying signals on graphs through their deviation from a stationary behavior: preliminary results on EEG signals –
*A. Dadras, Tabriz University; B. Pasdeloup, Ecole Polytechnique Fédérale de Lausanne* - Point Cloud Inpainting on Graphs from non-local Self-similarity –
*Z. Fu, Wei Hu and Z. Guo, Peking University*

### Oral session – GSP for Machine Learning 3 (chair: G. Leus)

List of papers- Learning Graph with Overlapping Community Structure –
*Y. Yuan, Singapore University of Technology and Design* - A Convex Hierarchical Clustering Perspective on Graph Scale Identification –
*C. Donnat, Stanford University* - Online Graph Learning from Sequential Data –
*S. Vlaski, H. Maretic, R. Nassif, P. Frossard and A. H. Sayed, Ecole Polytechnique Fédérale de Lausanne* - Uncertainty Quantification for Sampling-based Semi-Supervised Classification on Graphs –
*R. Varma and J. Kovacevic, Carnegie Mellon University*