Convolutional Neural Networks on Graphs
NTU
COM1 Level 2
SR3, COM1-02-12
Abstract:
Convolutional neural networks have greatly improved state-of-the-art performances in computer vision and speech analysis tasks, due to its high ability to extract multiple levels of representations of data. In this talk, we are interested in generalizing convolutional neural networks from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, telecommunication networks, or words' embedding. We present a formulation of convolutional neural networks on graphs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs. Numerical experiments demonstrate the ability of the system to learn local stationary features on graphs.
Biodata:
Xavier Bresson (PhD 2005, EPFL, Switzerland) is Associate Professor in Computer Science and member of the Data Science and AI Research Centre at NTU, Singapore. He is a leading researcher in the field of graph deep learning, a new framework that combines graph theory and deep learning techniques to tackle complex data domains in neuroscience, genetics, social science, physics, and natural language processing. He received in 2016 the highly competitive Singaporean NRF Fellowship of 2.5M US$ to develop these new techniques. He has organized international workshops and tutorials with Facebook, NYU, and USI about this emerging field such as the 2018 UCLA workshop, https://www.ipam.ucla.edu/programs/workshops/new-deep-learning-techniques, the 2017 CVPR tutorial, http://cvpr2017.thecvf.com/program/tutorials, and the 2017 NIPS tutorial, https://nips.cc/Conferences/2017/Schedule?showEvent=8735. He has published more than 60 peer-reviewed papers, including NIPS, ICML, JMLR, the top venues in machine learning. He was awarded several research grants in U.S. and Hong Kong. He has multiple consulting experiences with e.g. Nestle to design industrial deep learning techniques. On the teaching side, he was the main investigator in 2015 of the first course on deep learning at EPFL, Switzerland, which is now fully part of the EPFL undergraduate and graduate programs. He also designed a praised three-day data training on deep learning and standard techniques for various companies: http://data-science-training-xb.com.