Learning the Structure of EHR with Graph Convolutional Transformer - Edward Choi

Explore the potential of graph convolutional transformers in learning the structure of electronic health records (EHR) in this comprehensive lecture by Edward Choi. Delve into the graphical nature of EHR data stored in relational databases and discover how combining graph convolution with self-attention can enhance supervised prediction tasks. Gain insights into recent developments in multi-modal learning using Transformers and understand the application of interpretable deep learning methods for longitudinal electronic health records. Learn about hierarchical representations, regularization techniques, visit complexity, diagnosis code prediction, and the self-attention mechanism. Examine case studies, qualitative examples, and meaningful structures derived from learning the graphical structure of EHR data.

Introduction
What is EHR
Case Study
Structure of EHR
Hierarchical representation
Comparison
Regularization
Visit Complexity
Diagnosis Code Prediction
Selfattention mechanism
QKB operation
Learning the graphical structure
Learning meaningful structures
Qualitative examples
Conclusion