rllm.transforms.graph_transforms.GCNTransform

class rllm.transforms.graph_transforms.GCNTransform(normalize_features: str = 'l1')

Bases: GraphTransform

Preprocessing pipeline used by the original GCN model.

This transform is based on “Semi-supervised Classification with Graph Convolutional Networks” paper. GCNTransform applies a series of transformations to a graph, including:

1. Feature Normalization

2. Adjacency Matrix Normalization

   1. Adding Self-Loops

   2. Symmetric Normalization

Parameters:

normalize_features (str) – Feature normalization method passed to NormalizeFeatures. (default: "l1")