Daniel Worrall: Equivariance -Trends and Challenges
In the natural, world symmetries surround us in abundance and so it is of no surprise that we find them present in many datasets. This can be exploited when constructing deep models, such as classifiers or RL agents, by selecting only those operations/layers which preserve chosen symmetries. This design principle, known as equivariance, is the basis for the ubiquity of convolutions in computer vision and other areas. Following developments in the field of equivariance can be difficult due to the accelerated pace in recent years of both ideas and of the mathematical machinery used. In this talk I will collate some of the recent trends in the field and some open challenges that still exist.
Dr. Daniel Worrall is a machine learning researcher at Qualcomm AI Research, Amsterdam. Previously he was a postdoc in Max Welling’s group with the Amsterdam Machine Learning Lab (AMLAB) at the University of Amsterdam. He did his PhD at University College London in the Machine Vision Group with Gabriel J. Brostow and Dr. Clare Wilson FRCOphth. Before that he studied engineering at The University of Cambridge, and is a scholar of Sidney Sussex College. His current research interests include equivariance and combinatorial optimization, while previously he has worked on uncertainty quantification, unsupervised representation learning, variational inference, normalizing flows, and medical imaging.
Jelmer Wolterink: Learning on Graphs and Meshes for 3D Blood Vessel Analysis
In recent years, deep learning has had a tremendous impact on medical image analysis. Convolutional neural networks (CNNs) allow routine processing of medical images to extract e.g. high-quality triangular mesh segmentations of organs, or graphs representing the structure of vessel trees. A natural next step is to apply machine learning to these representations to extract additional clinically valuable information. However, current CNNs do not generalize to data that is not organized on a regular grid. In this talk, I will review how recent developments in graph and mesh convolutional networks can be leveraged to extract more structural, anatomical, and functional information from medical images, with applications in the improved analysis of 3D blood vessels.
Jelmer Wolterink is an assistant professor at the Department of Applied Mathematics of the University of Twente. He obtained his PhD in 2017 at Utrecht University (UMC Utrecht) with a thesis entitled Machine learning based analysis of cardiovascular images and was a postdoc in the quantitative image analysis group with Ivana Išgum at UMC Utrecht and Amsterdam UMC, Jelmer’s research interests are in novel machine learning techniques for medical image analysis, such as geometric deep learning and generative modeling. He is involved in several research projects, including an NWO VENI grant that allows him to develop geometric deep learning methods to model the progression of abdominal aortic aneurysms.
Marinka Zitnik: Graph Neural Networks for the Development of Therapeutics
Abstract: The success of machine learning depends heavily on the choice of representations used for downstream tasks. Graph neural networks have emerged as a predominant choice for learning representations of networked data. In this talk, I describe our efforts to expand the scope and ease the applicability of graph representation learning. First, I outline SubGNN, a subgraph neural network for learning disentangled subgraph representations. Second, I will describe G-Meta, a novel meta-learning approach for graphs. G-Meta uses subgraphs to generalize to completely new graphs and never-before-seen labels using only a handful of nodes or edges. G-Meta is theoretically justified and scales to orders of magnitude larger datasets than prior work. Finally, I will discuss applications in biology and medicine. The new methods have enabled the repurposing of drugs for emerging diseases, including COVID-19, where our predictions were experimentally verified in the wet laboratory. Further, the methods also allow for molecular phenotyping, much better than more complex algorithms. Lastly, I describe our efforts in learning actionable representations that allow users of our models to receive predictions that can be interpreted meaningfully.
Bio: Marinka Zitnik is an Assistant Professor at Harvard University with appointments in the Department of Biomedical Informatics, Broad Institute of MIT and Harvard, and Harvard Data Science. Dr. Zitnik investigates machine learning, focusing on challenges brought forward by interconnected data in science, medicine, and health. Her work received best paper and research awards from the International Society for Computational Biology and the Bayer Early Excellence in Science Award. She has recently been named a Rising Star in Electrical Engineering and Computer Science (EECS) by MIT and also a Next Generation in Biomedicine by the Broad Institute, being the only young scientist who received such recognition in both EECS and Biomedicine.
The event will held as a Zoom Webinar. Should you be interested in attending, please register here: