Module edge_sampling 

Edge sampling and train/test edge splits (ALGO-TR-009).

Utilities for randomly sampling edges and splitting a graph’s edge set into disjoint train/test partitions. Essential for link prediction evaluation: the test edges are held out, and the model must predict them.

Structs

EdgeSplit

Result of an edge train/test split.

Functions

sample_edges

Uniformly sample count edges from the graph without replacement.

split_edges

Split graph edges into train and test sets.

split_edges_connected

Split edges ensuring the training graph remains connected.