Physics-Inspired Learning on Graph - Michael Bronstein, PhD
Offered By: Open Data Science via YouTube
Course Description
Overview
Explore physics-inspired learning on graphs in this 27-minute video featuring AI expert Michael Bronstein. Delve into the limitations of current graph deep learning schemes and discover a revolutionary approach that challenges the prevailing "node-and-edge"-centric mindset. Learn about continuous learning models inspired by differential geometry, algebraic topology, and differential equations. Gain insights into topics such as geometric deep learning, graph neural networks, diffusion equations, graph rewiring, Beltrami flow, Ricci flow, and cellular sheaves. Understand why message passing may no longer be sufficient for advancing graph machine learning and explore new avenues that could revolutionize the field.
Syllabus
- Introductions & Opening
- Geometric Deep Learning
- Graph Neural Networks
- Diffusion Equation
- Spatial Derivative: Graph Rewiring?
- Beltrami Flow
- Graph Deltrami Flow
- Ricci Flow
- Cellular Sheaves
- Are We Done With Message Pasing?
Taught by
Open Data Science
Related Courses
Graph Attention Networks - GNN Paper ExplainedAleksa Gordić - The AI Epiphany via YouTube Geometric Deep Learning for Drug Discovery
IEEE Signal Processing Society via YouTube Detection of Objects in Cryo-Electron Micrographs Using Geometric Deep Learning
Institute for Pure & Applied Mathematics (IPAM) via YouTube Inverse Problems on Graphs with Geometric Deep Learning
APS Physics via YouTube A Sheaf-based Approach to Graph Neural Networks
Applied Algebraic Topology Network via YouTube