Learn about the use cases of graph modeling and find out how to train graph neural networks and evaluate its results.
Overview
Syllabus
Introduction
- Introducing graph neural networks
- Prerequisites
- Undirected and directed graphs
- Other graph types
- Graph representations
- Prediction tasks with graphs
- Approaches to graph machine learning
- Challenges of using graphs in machine learning
- Graph neural networks intuition
- Understanding the structure of GNNs
- The graph neural network architecture
- Message passing transformation and aggregation
- Training a GNN
- Introducing PyTorch Geometric
- Exercise: Set up the Colab environment and libraries
- Exercise: Setting up a graph data structure in PyG
- Exercise: Visualizing graphs and exploring graph methods
- Exercise: Visualizing and exploring a directed graph
- Exercise: Exploring the cora dataset
- Exercise: Mini batches of data
- Exercise: Representing heterogeneous graphs in PyG
- Exercise: The CiteSeer dataset for node classification
- Exercise: Setting up a DNN as a baseline model
- Exercise: Training the baseline model
- Exercise: Setting up a graph convolutional network
- Exercise: Training a GCN
- Summary and next steps
Taught by
Janani Ravi