Collaborative Research Unit

Group for Artificial Intelligence and Sparse Modelling

Acronym
GAIM
Other information
Research disciplines
  • Engineering and technology
    • Biomedical image processing
    • Computer vision
    • Image and language processing
    • Interactive and intelligent systems
    • Pattern recognition and neural networks
    • Data visualisation and imaging
Description
Research Group for Artificial Intelligence and Sparse Modelling (GAIM) is part of the Department of Telecommunications and Information Processing at Ghent University. GAIM’s research is at the intersection of machine learning, signal processing and information theory. We pursue the development and integration of innovative algorithms for the analysis of high-dimensional data, including pattern recognition, classification and anomaly detection in multimodal data, recovery of signals and images fromincomplete and corrupted data and making inferences under uncertainty. The underlying theoretical concepts include searching for compact data representations (sparse coding), deep learning, statistical modelling and Bayesian reasoning. GAIM has extensive expertise in hierarchical signal and image representations (including wavelet representation and extensions), sparse coding (including dictionary learning), compressed sensing, statistical image modelling, probabilistic graphical models such as Markov Random Fields and in Bayesian detection and estimation in general. The application areas of this research include machine vision, biomedical processing, remote sensing, computer graphics and art investigation. The group has built strong international reputation in high-dimensional data processing and analysis. Some of our most important results are in the field of restoration, classification and anomaly detection in multimodal and hyperspectral images, compressed sensing of medical images and analysis of master paintings. Our current research focuses largely on classification and clustering of large-scale data, neural architecture search (including optimisation of convolutional neural networks), deep generative models (variational autoencoders) and geometric deep learning (extending deep learning models for non-Euclidean data, such as those represented by graphs and meshes). Regarding education, GAIM gives currently three Master courses at the Faculty of Engineering and Architecture: Artificial Intelligence, Computer Graphics and Probabilistic Graphical Models. GAIM is a member of the UGent.AI initiative (within the Flanders AI Research Program of the Flemish Government) and member of the HyCT and iKNOW valorisation consortia. Regarding wider national and international context, GAIM coordinates a UGent-VUB Alliance Research Group International Big Data analytics Lab (iBDL), which is part of the International Research Group on Big Data (between Ghent University, Vrije Universiteit Brussel, University College London and Duke University). GAIM is also a founding member of the international research network Turning Images into Value through Statistical Parameter Estimation, and co-organiser of reputed iTWIST series of workshops (on interactions between low-complexity data models such as sparse or low-rank data models and novel data sensing techniques). GAIM plays an important role in the area of information processing for art investigation, among others as a co-organizer of the well-established IP4AI workshops (Image Processing for Art Investigation). Finally, next to its involvement in IEEE committees, GAIM contributes actively to the European Section of IEICE promoting the collaboration between Japan and Europe,through its continued involvement over many years in the organization of the IEICE Information and Technology Forum.