SHD

class pgmpy.metrics.SHD

Bases: _BaseSupervisedMetric

Computes the Structural Hamming Distance between true_causal_graph and est_causal_graph.

SHD is defined as total number of basic operations: adding edges, removing edges, and reversing edges required to transform one graph to the other. It is a symmetrical measure.

The code first accounts for edges that need to be deleted (from true_model), added (to true_model) and finally edges that need to be reversed. All operations count as 1.

Examples

>>> from pgmpy.metrics import SHD
>>> from pgmpy.base import DAG
>>> dag1 = DAG([(1, 2), (2, 3)])
>>> dag2 = DAG([(2, 1), (2, 3)])
>>> shd = SHD()
>>> shd(true_causal_graph=dag1, est_causal_graph=dag2)
1