Speaker: Prof. A.G. Marques (King Juan Carlos University, Spain) 

Time and Date: 11:00-12:00 am, July 02, 2018

Place: Room B415 of Computing Center Building, Handan Campus, Fudan University

Abstract:

Convolut10nal neural networks (CNNs) are being applied to an mcreasmg number of problems and fields due to their superior performance in classification and regression tasks. Motivated by the recent interest in processing signals defined in irregular domains, we investigate CNN architectures that operate on signals supported on graphs. This is challenging problem since two of the key operations that CNNs implement are convolution and pooling, which are implicitly defined to act on regular domains. In this talk, we describe two architectures that generalize CNNs for the processing of signals supported on graphs. The selection Graph Neural Network (GNN) repl妇ces linear time invariant filters with linear shift invariant graph filters to generate convolutional features and reinterprets pooling as a possibly nonlmear subsampling stage where nearby nodes pool their information in a set of preselected sample nodes. A key component of the architecture 1s to remember the position of sampled nodes to permit computation of convolutional features at deeper layers. The aggregation GNN diffuses the signal through the graph and stores the sequence of diffused components observed by a designated node. This procedure effectively aggregates all components into a stream of information having temporal structure to which the convolution and pooling stages of regular CNNs can be applied. A multi-node version of aggregation CNNs is further introduced for operation in large scale graphs. An important property of selection and aggregation GNNs is that they reduce to conventional CNNs when particularized to time signals _ remterpreted as graph signals in a circulant gral?h·C?mparative numerical analyses are performed in a synthetic source localization application. Performance is evaluated for a text category class1ficat10n problem using word proximity networks. 

Biography:

Antonio G. Marques received the telecommunications engineering degree and the Doctorate degree, both with highest honors, from the Carlos III University of Madrid, Madrid, Spain, in 2002 and 2007, respectively. In 2007, he became a faculty in the Department of Signal The?ry and Communications, King Juan Carlos University, Madrid, Spain, where he currently develops his research and teaching activities as an Associate Professor. From 2005 to 2015, he held different visiting positions at the University of Minnesota, Minneapolis, MN, USA. In 2015 and 2016, he was a Visitor Scholar in the University of Pennsylvania, Philadelphia, PA, USA. 
His research interests lie in the areas of signal processing, networking and communications. His current research focuses on stochastic optimization of wireless and power networks, signal processing for graphs, and nonlinear network optimization. He has served the IEEE in a number of posts, collaborating on the organization of more than 20 IEEE conferences and workshops. Currently, he is an Associate Editor of the SIGNAL PROCESS­ING LETTERS, a member of the IEEE Signal Processing Theory and Methods Technical Committee and a member of the IEEE Signal Processing for Big Data Special Interest Group. Dr. Marques' work has been awarded in several conferences and workshops, with recent best paper awards including Asilomar 2015, IEEE SSP 2016 and IEEE SAM 2016