rllm.transforms.graph_transforms.RECTTransform¶

class rllm.transforms.graph_transforms.RECTTransform(normalize_features: str = 'l1', svd_out_dim: int = 200, use_gdc: bool = False)[source]¶

Bases: GraphTransform

The RECTTransform class is based on the method described in the “Network Embedding with Completely-imbalanced Labels” paper. This transform applies a series of transformations to a graph, including feature normalization, reduce the dimensionality of the features and adjacency matrix normalization.

Parameters:

normalize_features (str) – Method for feature normalization. (default: "l1")
svd_out_dim (int) – The output dimensionality after SVD feature reduction. (default: 200)
use_gdc (bool) – Whether to use Graph Diffusion Convolution (GDC) instead of GCN normalization. (default: False)