Learning And Signal Processing

Our research activity focused on developing novel adaptive strategies for large-scale networks with applications including adaptive streaming strategies for virtual reality services, data-efficient multi-arm bandit problems for online recommendation systems, graph-based reinforcement learning for AI systems, and influence maximization over social networks. Our research is at the crossroad between multimedia processing, machine learning, and signal processing.

UCL Christmas Party 2018

Research News

Laura co-presenting the ICASSP 2021 Tutorial on “Graph signal processing for machine learning”

Title: “Graph signal processing for machine learning: A review and new perspectives”Presenters: Xiaowen Dong, Dorina Thanou, Laura Toni, Michael Bronstein, and Pascal Frossard Description: The effective representation, processing, analysis, and visualization of large-scale structured data, especially those related to complex domains such as networks and graphs, are one of the key questions in modern machine learning. Graph signal processing (GSP),Continue Reading

Laura co-organizing the 1st Workshop “When Graph Signal Processing meets Computer”, ICCV 2021

Workshop Title: “When Graph Signal Processing meets Computer Vision”Organizers: T. Bouwmans, G. Cheung, W. Hu, Y. Tanaka, L. Toni Link: https://gsp-cv.univ-lr.fr/gspcv-21/ Graph signal processing (GSP)  is the study of computational tools to process and analyze data residing on irregular correlation structures described by graphs. Early GSP researchers explored low-dimensional representations of high-dimensional data via spectral graph theory—mathematical analysis of eigen-structures of theContinue Reading

Paper Accepted to IEEE International Symposium on Information Theory (ISIT) 2021

Title: “Information-Theoretic Bounds on the Moments of the Generalization Error of Learning Algorithms” Authors: Gholamali Aminian, Laura Toni, Miguel R. D. Rodrigues Link: https://arxiv.org/abs/2102.02016 Abstract: Generalization error bounds are critical to under- standing the performance of machine learning models. In this work, building upon a new bound of the expected value of an arbitrary function of the population and empiricalContinue Reading

Check our opening page, there is always an opportunity if you want to join us 🙂