Expert Talks


Expert Talk Speaker Timing Overview
ET1: Attention and Transformers M. J. Zaki 10:00 AM to 11:30 AM Self-attention is arguably the current driving force behind most advances in deep learning from text to graphs and vision. We will cover the main ideas behind attention and the Transformer architecture, with applications in language modeling and computer vision.
ET2: Graph Neural Networks M. J. Zaki 11:45 AM to 01:15 PM Heterogeneous graphs can model most real world datasets ranging from text, to images, proteins, knowledge graphs, and so on. We will cover the fundamentals of graph neural networks, including graph convolutions and graph attention.
ET3: Advanced Deep Learning M. J. Zaki 02:30 PM to 04:00 PM We will cover some advanced forms of attention mechanisms, such as axial attention, structural attention, and visual transformers, with applications in vision, protein structure prediction (DeepMind's Alphafold), and so on.
ET4: Analyzing Clinical Narratives for Better Health Care Management L. Dey 04:00 PM to 05:30 PM Clinical narratives in the form of patient health records, physician and nurses’ notes, imaging and test reports contain a wealth of information that can significantly enhance the quality of predictive and prescriptive analytics for health care management. According to some reliable sources, the volume of untapped clinical text data runs into several thousand exabytes. In this talk we will explore how NLP techniques can be used to dig into this rich source of information and extract valuable insights, that can significantly improve the state of the art in health care analytics.