Michael Felsberg received the Ph.D. degree in engineering from the University of Kiel, Kiel, Germany, in 2002. Since 2008, he has been a Full Professor and the Head of the Computer Vision Laboratory, Linköping University, Sweden. His current research interests include signal processing methods for image analysis, computer and robot vision, and machine learning. He has published more than 100 reviewed conference papers, journal articles, and book contributions. He was a recipient of awards from the German Pattern Recognition Society in 2000, 2004, and 2005, from the Swedish Society for Automated Image Analysis in 2007 and 2010, from Conference on Information Fusion in 2011 (Honorable Mention), and from the CVPR Workshop on Mobile Vision 2014. He has achieved top ranks on various challenges (VOT: 3rd 2013, 1st 2014, 2nd 2015; VOT-TIR: 1st 2015; OpenCV Tracking: 1st 2015; KITTI Stereo Odometry: 1st 2015, March). He has coordinated the EU projects COSPAL and DIPLECS, he is an Associate Editor of the Journal of Mathematical Imaging and Vision, Journal of Image and Vision Computing, Journal of Real-Time Image Processing, Frontiers in Robotics and AI. He was Publication Chair of the International Conference on Pattern Recognition 2014 and Track Chair 2016, he was the General Co-Chair of the DAGM symposium in 2011, and he will be general Chair of CAIP 2017.

Title: 

From conformal modeling to O(n) equivariant deep learning

Abstract: Felix Klein’s Erlangen Programme of 1872 introduced a methodology to unify non-Euclidean geometries. Within machine learning, geometric deep learning (GDL) constitutes a unifying framework for various neural network architectures and allows models to adjust to different geometric transformations. The orthogonal group O(n) fully encapsulates the symmetry structure of an nD sphere, including both rotational and reflection symmetries. Neurons with spherical decision surfaces —spherical neurons — can efficiently be constructed using a conformal embedding of Euclidean space and their activation transforms in an equivariant way under rotations. Replicating spherical neurons at the vertices of the regular simplex results in the steerable TetraSphere representation, useful in various applications, e.g. on pointclouds.

Bram van Ginneken was born in Nuenen in 1970. He studied Physics at Eindhoven University of Technology and Utrecht University. In 2001, he obtained his PhD at the Image Sciences Institute on Computer-Aided Diagnosis in Chest Radiography, where he continued to work and set up a research group on medical image analysis. His PhD research on automated detection of tuberculosis resulted in medical device software called CAD4TB. With installations in over 75 countries worldwide, CAD4TB is the most widely used autonomous AI solution for the interpretation of medical images. In 2010, he moved to Radboud University Medical Center where he set up the Diagnostic Image Analysis Group and was appointed full professor in 2012. He (co-)authored over 300 publications in international journals. Since 2010, he also works for the Fraunhofer Institute for Digital Medicine MEVIS in Bremen, Germany. In 2014, he founded Thirona, a company that develops software for CT lung image analysis. He pioneered the concept of challenges in medical image analysis and created grand-challenge.org. In 2024, he founded Plain Medical, a company that develops AI solutions to reduce the workload of radiologists.

Title: 

Will AI breakthroughs solve the healthcare crisis?  

Abstract: My talk consists of 5 parts. First, I will explain what the healthcare crisis is. Surprisingly, it has nothing to do with what nearly all research in the field is focused on: trying to make healthcare better. Next, I will discuss the major breakthroughs in AI. First, solving computer vision, illustrated with healthcare applications of automated image analysis in radiology and pathology. Second, the large transformer networks that allow computers to read and write. I will show its potential with an application in emergency care medicine. Third, the breakthrough that is not realized yet: capable robots that can perform the simple tasks that comprise the bulk of human labor. The final part zooms out and discusses the possible consequences of these developments.

Marie-Francine (Sien) Moens holds a MSc ("licentiaat") in Computer Science and a PhD degree in Computer Science from KU Leuven. She is holder of the ERC Advanced Grant CALCULUS (2018-2024) granted by the European Research Council. From 2012 till 2016 she was the coordinator of the MUSE project financed by Future and Emerging Technologies (FET) - Open of the European Commission. In 2021 she was the general chair of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP 2021). In 2011 and 2012 she was appointed as chair of the European Chapter of the Association for Computational Linguistics (EACL) and was a member of the executive board of the Association for Computational Linguistics (ACL). She is currently associate editor of the journal IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) and was a member of the editorial board of the journal Foundations and Trends® in Information Retrieval from 2014 till 2018. From 2010 until 2014 she was a member of the Research Council of KU Leuven and from 2014 until 2018 she was a member of the Council of the Industrial Research Fund of KU Leuven. From 2014 till 2018 she was the scientific manager of the EU COST action iV&L Net (The European Network on Integrating Vision and Language). In 2014 she was a Scottish Informatics and Computer Science Alliance (SICSA) Distinguished Visiting Fellow. She is a fellow of the European Laboratory for Learning and Intelligent Systems (ELLIS).

Her research topics include among others: Machine learning for natural language processing and the joint processing of language and visual data; representation learning for language grounding in the physical and social world; Deep learning, latent variable models and brain inspired models; Machine learning models for structured prediction and generation.

Title: 

Deep Learning for Natural Language Understanding: The Tension Between Contextuality, Sparsity and Identity

Abstract: We identify several challenges faced by current natural language understanding models, particularly in comprehending lengthy discourses such as narratives and in tasks requiring reasoning with content. Many of these issues are equally significant in the processing of visual content and multimodal tasks, such as video understanding or text-to-video generation. To address these challenges, we compare how machines process content with what is known about the language and multimodal processing mechanisms of the human brain, focusing specifically on the human default mode network. This comparison leads to potential solutions for improving machine understanding, including modifications to model architectures and training methods. These modifications emphasize representations that model contextuality, sparsity, and identity, thereby enhancing compositionality, controllability, and explainability. Finally, we highlight applications where such advancements in representation learning could have a significant impact.

Line Clemmensen is a Professor at the University of Copenhagen. Her research focus is statistical modelling and machine learning and statistical evaluation of machine learning and AI. I have special interests in low resource domains, explainable and fair modelling within health and life science applications. 

Title: 

Fair AI in Psychiatry

Abstract: This talk demonstrates a range of applications of AI within psychiatry and discuss ML fairness in this context. Examples include predictions of OCD events using wearables, behavioral coding using videos, and predictions of suicide and suicide attempt using health registers.