Graph Neural Networks Are Dynamic Programmers

Explore the intersection of graph neural networks and dynamic programming in this IEEE Signal Processing Society webinar. Delve into classical algorithms, optimal path finding, and neural approaches as Petar Veličković from Deepmind guides you through data efficiency, end-to-end pipelines, and algorithmic reasoning. Examine mathematical preparations, data flow diagrams, and practical illustrations to understand how graph neural networks function as dynamic programmers. Gain insights into this cutting-edge topic and its applications in data science and signal processing.

Introduction
Classical Algorithms
Overview
Motivation
Optimal Path
Neural Algorithms
Data Efficiency
EndtoEnd Pipeline
Algorithmic Reasoning
General Idea
Summary
Graph Neural Networks
Dynamic Programming
Dynamic Programming Example
Mathematical Preparation
Illustration
Aggregation
Data Flow Diagram
Conclusion
Questions