Representation Learning on Graphs and Networks

Explore the cutting-edge realm of graph representation learning in this 26-minute talk by Dr. Petar Veličković. Delve into the surge in research on deep graph embeddings, extensions of CNNs to graph-structured data, and neural message-passing approaches revolutionizing domains like chemical synthesis, vehicle routing, 3D-vision, and recommender systems. Gain insights into permutation invariance and equivariance principles, understanding how graph neural networks (GNNs) transcend traditional input graph structures. Learn about the ubiquity of graphs and GNNs, the rich ecosystem of libraries supporting this field, and the important constraint of locality in GNN layers. Discover the three flavors of GNNs and their applications, equipping yourself with knowledge at the forefront of AI and data science innovation.

- Introductions
- Graphs are everywhere!
- Graph Neural Networks are Everywhere!
- Why are we here?
- Rich ecosystem of libraries
- Permutation Invariance and Eqauivariance
- Important Constraint: Locality
- What’s in a GNN layer?
- If you’d like to know more about the three flavors…