rllm.transforms

Graph Transforms

Basic Level

NodeTransform	Base class for node-wise transformations on graph data.
EdgeTransform	Base class for edge-wise transformations on graph data.
NormalizeFeatures	Row-normalizes the node features.
SVDFeatureReduction	Dimensionality reduction of node features via Singular Value Decomposition (SVD).
KNNGraph	Builds a k-NN adjacency matrix from node features.
AddRemainingSelfLoops	Adds missing self-loops to the adjacency matrix.
RemoveSelfLoops	Removes diagonal self-loop edges from an adjacency matrix.
GCNNorm	Applies the standard GCN adjacency normalization.
GDC	Processes the graph via Graph Diffusion Convolution (GDC) from the "Diffusion Improves Graph Learning" paper (functional name: gdc).

Model Level

GraphTransform	The GraphTransform class is a base class for applying a series of transformations to graph data.
GCNTransform	Preprocessing pipeline used by the original GCN model.
RECTTransform	The RECTTransform class is based on the method described in the "Network Embedding with Completely-imbalanced Labels" paper.

Table Transforms

Basic Level

ColTransform	An abstract base class for transforming individual column features in table data.
ColNormalize	The ColNormalize class is designed to normalize numerical features in tabular data.
OneHotTransform	One-hot encodes categorical features.
StackNumerical	The StackNumerical class is designed to transform numerical features in tabular data by stacking them into a specified dimension.

Model Level

TableTransform	Base class for table Transform.
DefaultTableTransform	Default table transform that only performs missing-value handling.
TabTransformerTransform	TabTransformerTransform applies ColNormalize transform to tabular data specifically for the TabTransformer model.

Utils

RemoveTrainingClasses

Removes classes from the node-level training set as given by data.train_mask, e.g., in order to get a zero-shot label scenario.