Graph Classification

Whole-graph classification with pooling methods on TU datasets and MNIST superpixels.

All categories

---

• Counting colors with supervised Top-k Pooling (COLORS-3)

  Count how many nodes of a given color a graph contains.

  COLORS-3 (TUDataset) · GINConv · TopKPooling

  Read Tutorial    

• Counting triangles with supervised SAGPool (TRIANGLES)

  Count the triangles in a graph.

  TRIANGLES (TUDataset) · GCNConv · GINConv

  Read Tutorial    

• DMoN Pooling on PROTEINS

  Classify proteins as enzymes or non-enzymes.

  PROTEINS (TUDataset) · DenseGraphConv · GCNConv

  Read Tutorial    

• DiffPool on PROTEINS

  Classify proteins as enzymes or non-enzymes.

  PROTEINS (TUDataset) · DenseSAGEConv

  Read Tutorial    

• Graph Isomorphism Network (GIN) on MUTAG

  Predict whether a molecule is mutagenic.

  MUTAG (TUDataset) · GINConv · MLP

  Read Tutorial    

• Graph Multiset Transformer (GMT) on PROTEINS

  Classify proteins as enzymes or non-enzymes.

  PROTEINS (TUDataset) · GCNConv · GraphMultisetTransformer

  Read Tutorial    

• Graph classification with a root-node readout (UPFD model)

  PyG's UPFD example detects fake news: each graph is a news propagation tree whose root node is the news article itself.

  PROTEINS (TUDataset) · GCNConv

  Read Tutorial    

• Graph classification with the Weisfeiler-Lehman kernel (ENZYMES)

  Classify enzymes into 6 classes without training a neural network.

  ENZYMES (TUDataset) · WLConv

  Read Tutorial    

• Image classification with NNConv and Graclus pooling

  Classify handwritten digits represented as graphs.

  MNISTSuperpixels · NNConv

  Read Tutorial    

• Image classification with SplineConv and Graclus pooling

  Classify handwritten digits represented as graphs.

  MNISTSuperpixels · SplineConv

  Read Tutorial    

• Image classification with SplineConv and voxel-grid pooling

  Classify handwritten digits represented as graphs.

  MNISTSuperpixels · SplineConv

  Read Tutorial    

• Memory-based Graph Pooling (MemPool) on PROTEINS

  Classify proteins as enzymes or non-enzymes (PROTEINS_full, which includes extra node attributes).

  PROTEINS_full (TUDataset) · GATConv · MemPooling

  Read Tutorial    

• MinCut Pooling on PROTEINS

  Classify proteins as enzymes or non-enzymes.

  PROTEINS (TUDataset) · DenseGraphConv · GCNConv

  Read Tutorial    

• Top-k Pooling on PROTEINS

  Classify proteins as enzymes or non-enzymes.

  PROTEINS (TUDataset) · GraphConv · TopKPooling

  Read Tutorial