Code Examples

Welcome to the K3-Node Code Examples. Browse by category, or jump straight to a notebook below.

Categories

• Getting Started — 1 example
• Clustering — 2 examples
• Graph Classification — 14 examples
• Inductive Learning — 3 examples
• Knowledge Graphs — 6 examples
• Large-Scale & Scalable Training — 10 examples
• Link Prediction — 8 examples
• Molecular Property Prediction — 6 examples
• Node Classification — 24 examples
• Point Cloud & 3D — 9 examples
• Representation Learning — 6 examples
• Utilities & Misc — 4 examples

All Examples

• A GNN made heterogeneous, on a movie graph — Knowledge Graphs
• A simple GCN baseline for node property prediction — Large-Scale & Scalable Training
• ARMA Graph Convolutions on Cora — Node Classification
• Attention-based GNN (AGNN) on Cora — Node Classification
• Classifying nodes from their own subgraphs (ShaDow-GNN) — Large-Scale & Scalable Training
• Correct and Smooth on Cora — Node Classification
• Counting colors with supervised Top-k Pooling (COLORS-3) — Graph Classification
• Counting triangles with supervised SAGPool (TRIANGLES) — Graph Classification
• DMoN Pooling on PROTEINS — Graph Classification
• DiffPool on PROTEINS — Graph Classification
• Directed GNN (Dir-GNN) on WebKB — Node Classification
• Distilling a GNN into an MLP (GLNN) — Utilities & Misc
• Dynamic Neighborhood Aggregation (DNA) on Cora — Node Classification
• Efficient Graph Convolution (EGC) on BACE molecules — Node Classification
• Entity classification with R-GCN (AIFB) — Knowledge Graphs
• Entity classification with relational graph attention (AIFB) — Knowledge Graphs
• Faster neighbor sampling with hierarchical trimming — Large-Scale & Scalable Training
• Feature-wise Linear Modulation (FiLM) on PPI — Node Classification
• Forecasting events in a temporal knowledge graph (RE-Net) — Link Prediction
• GCNII: a 64-layer GCN on Cora — Node Classification
• GeniePath on PPI — Node Classification
• Graph Attention Network (GAT) on Cora — Node Classification
• Graph Convolutional Network (GCN) on Cora — Node Classification
• Graph Isomorphism Network (GIN) on MUTAG — Graph Classification
• Graph Multiset Transformer (GMT) on PROTEINS — Graph Classification
• Graph U-Net on Cora — Node Classification
• Graph classification with a root-node readout (UPFD model) — Graph Classification
• Graph classification with the Weisfeiler-Lehman kernel (ENZYMES) — Graph Classification
• High-level task APIs — Getting Started
• Image classification with NNConv and Graclus pooling — Graph Classification
• Image classification with SplineConv and Graclus pooling — Graph Classification
• Image classification with SplineConv and voxel-grid pooling — Graph Classification
• Inductive Deep Graph Infomax with GraphSAGE and neighbor sampling — Representation Learning
• Knowledge graph completion with R-GCN and DistMult — Knowledge Graphs
• Knowledge graph embeddings: TransE, DistMult, ComplEx and RotatE — Knowledge Graphs
• LINKX on WebKB — Node Classification
• Label Propagation on Cora — Node Classification
• Learning the 2nd smallest number with LCM Aggregation — Node Classification
• Learning the median with Equilibrium Aggregation — Node Classification
• Link prediction as graph classification (SEAL) — Link Prediction
• Link prediction with LPFormer (Cora) — Link Prediction
• Link prediction with a GCN (Cora) — Link Prediction
• Link prediction with attract-repel embeddings (Cora) — Link Prediction
• Link prediction with graph autoencoders (GAE and VGAE) — Link Prediction
• Logging training to TensorBoard — Utilities & Misc
• Memory-based Graph Pooling (MemPool) on PROTEINS — Graph Classification
• Memory-efficient deep GNNs with reversible layers (RevGNN) — Large-Scale & Scalable Training
• MinCut Pooling on PROTEINS — Graph Classification
• Mini-batch training of GNNs and graph transformers — Large-Scale & Scalable Training
• Mini-batch training on graph clusters (Cluster-GCN) — Large-Scale & Scalable Training
• Mini-batch training on sampled subgraphs (GraphSAINT) — Large-Scale & Scalable Training
• Mini-batch training with neighbor sampling (GraphSAGE) — Large-Scale & Scalable Training
• MixHop on Cora — Node Classification
• Multi-label protein function prediction with Cluster-GCN (PPI) — Inductive Learning
• Multi-label protein function prediction with GAT (PPI) — Inductive Learning
• Multi-label protein function prediction with a deep GCNII (PPI) — Inductive Learning
• Node clustering with an adversarially regularized autoencoder (ARGVA) — Clustering
• Node embeddings from random walks (Node2Vec) — Representation Learning
• Point cloud classification with DGCNN — Point Cloud & 3D
• Point cloud classification with Point Transformer — Point Cloud & 3D
• Point cloud classification with PointNet++ — Point Cloud & 3D
• Point cloud classification with RandLA-Net — Point Cloud & 3D
• Point cloud segmentation with DGCNN — Point Cloud & 3D
• Point cloud segmentation with Point Transformer — Point Cloud & 3D
• Point cloud segmentation with PointNet++ — Point Cloud & 3D
• Point cloud segmentation with RandLA-Net — Point Cloud & 3D
• Pre-trained DimeNet models on QM9 — Molecular Property Prediction
• Pre-trained SchNet models on QM9 — Molecular Property Prediction
• Pre-trained positional and structural encodings (GPSE) — Representation Learning
• Predicting future interactions with Temporal Graph Networks (TGN) — Link Prediction
• Predicting quantum properties of molecules with an MPNN (QM9) — Molecular Property Prediction
• Predicting solubility with AttentiveFP (ESOL) — Molecular Property Prediction
• Predicting solubility with Principal Neighbourhood Aggregation — Molecular Property Prediction
• Predicting solubility with a graph transformer (GraphGPS) — Molecular Property Prediction
• Predicting trust and distrust with a signed GCN (Bitcoin-OTC) — Link Prediction
• Propagational MLP (PMLP) on Cora — Node Classification
• Recommendation with LightGCN (MovieLens) — Utilities & Misc
• Relational deep learning with a heterogeneous GraphSAGE — Knowledge Graphs
• Scalable Inception Graph Networks (SIGN) on Cora — Node Classification
• Semi-supervised node classification with OGC — Clustering
• Shape correspondence with SplineCNN — Point Cloud & 3D
• Simple Graph Convolution (SGC) on Cora — Node Classification
• Spline Convolutions on Cora — Node Classification
• Streaming data pipelines for graphs — Utilities & Misc
• SuperGAT on Cora — Node Classification
• Top-k Pooling on PROTEINS — Graph Classification
• Topology Adaptive GCN (TAGCN) on Cora — Node Classification
• Unsupervised GraphSAGE (Cora) — Representation Learning
• Unsupervised GraphSAGE on protein graphs (PPI) — Representation Learning
• Unsupervised node embeddings with Deep Graph Infomax (Cora) — Representation Learning
• Using known labels as features (UniMP) — Large-Scale & Scalable Training
• Very deep GNNs for protein function prediction (DeeperGCN) — Large-Scale & Scalable Training
• Zero-shot node classification with RECT on Cora — Node Classification