PyTorch Geometric for Graph Neural Nets
summary
Live code demo builds a 3-layer GCNConv model in PyTorch Geometric that maps 3-d node features to a 64-d embedding and uses global mean pooling to predict 6 classes, with slides and code linked. Uses the ENZYMES dataset (example graph: 37 nodes, 168 edges) and shows batching as one big graph, degree-based normalization, adding self-loops, and a custom MessagePassing subclass for bespoke message, aggregate, and update functions.
AI-written summary · about
chapters (22)
- 0:00Graph neural networks intro
- 1:33Deep learning to graph data
- 5:07Graphs non-Euclidean data
- 6:10Graph examples and tasks
- 8:17Graphs with adjacency matrices
- 9:21Graph neural networks and their outputs
- 11:34Message passing in GNNs
- 16:31Message passing variations
- 19:04Attention in graph nets
- 21:56Pooling for graph processing
- 23:00PyTorch Geometric basics
- 24:48Batching graphs in PyG
- 27:39Simple GNN with PyG
- 32:07Training a graph neural network
- 32:29Custom GNN layers
- 34:54Closing remarks
- 35:17Q&A: Two-layer GNN
- 39:07Q&A: Unsupervised GNN models
- 40:56Q&A: Graph autoencoders
- 41:21Q&A: Node-specific weights
- 42:26Q&A: Variable graph sizes
- 44:57Q&A: Normalization in GNNs
related talks
59mTransforming DataFrames: Intro to Graphs & GNNs with PyTorch Geometric
31mGraph-Based Machine Learning with Python
29mGraph Neural Networks + LLM on
01h 30mPractical graph neural networks in Python with TensorFlow and Spektral
39mUsing Graph Neural Networks to Embrace the Dependency Within Your Data
18mTC-GNN: Bridging Sparse GNN Computation and Dense Tensor Cores on GPUs
16mPreview: Graph NNs
01h 23mGraph Neural Networks in Drug Discovery