Dive deep into Graph Neural Networks (GNN) implementation with Python in this comprehensive tutorial. Learn technical details and gain hands-on experience using NetworkX, PyG (pytorch_geometric), and matplotlib libraries. Explore graph representations, node embedding, message passing, and GNN explainers. Practice node classification using MLP and GNN, visualize graphs with NetworkX and tSNE, and understand various graph datasets available in PyG. Master concepts like adjacency matrices, convolution on graphs, and message passing through practical Jupyter notebook exercises and a detailed PDF presentation. Ideal for those seeking to enhance their understanding of GNN and its applications in machine learning and data analysis.
Overview
Syllabus
- Video Starts
- Video Introduction
- Tutorial Content in Part2
- Graph Representations Techniques
- Adjacency Matrix
- Incidence Matrix
- Degree Matrix
- Laplacian Matrix
- Creating Graph with NetworkX Jupyter notebook
- Graph Visualization with Node classes Jupyter notebook
- Graph Visualization with Node and Edge Labels Jupyter notebook
- Nodes Adjacency List Jupyter notebook
- Bag of Nodes
- Graph Walking Jupyter notebook
- GNN Concepts
- Role of Laplacian Matrix
- Convolution in Images
- Graph vs 2D fixed data types i.e. images, text
- Convolution on Graphs, how?
- Graph Feature Matrix
- Applying Convolution in Graphs
- Node Embeddings
- Message Passing in GNN
- Advantages of Node Embeddings
- GNN Use Cases
- Handling data in PyG Jupyter notebook
- GNN Experiment for Node grouping Jupyter notebook
- Node assignment to proper class Jupyter notebook
- GNN Model visualization with Netron
- Node classification using GNN in PyG
- Graph tSNE Visualization
- GNN Explainer
- Recap
Taught by
Prodramp
