pyagc.metrics.structure_metrics

structure_metrics(edge_index: Tensor, clusters: Tensor, metrics: Union[str, Tuple[str, ...]] = ('Mod', 'Cond'), vectorized: bool = True) → Dict[str, float]

Computes structural clustering metrics on a graph.

Supports modularity and conductance measures for evaluating the structural quality of clustering on graphs.

Parameters:

• edge_index (torch.Tensor) – Undirected edge index tensor of shape (2, num_edges).

• clusters (torch.Tensor) – Cluster assignments of shape (num_nodes,).

• metrics (str or tuple of str, optional) – Structural metrics to compute. Valid options are: 'Mod', 'Cond'. (default: ('Mod', 'Cond'))

• vectorized (bool, optional) – Whether to use vectorized computation for better performance. (default: True)

Returns:

Dict[str, float] – Dictionary mapping metric names to their computed values.

Example

>>> # Fast vectorized version (default)
>>> result = structure_metrics(edge_index, clusters, metrics=('Mod', 'Cond'))
>>> print(result)
{'Mod': 0.45, 'Cond': 0.32}

>>> # Loop-based version
>>> result = structure_metrics(edge_index, clusters, vectorized=False)