def calc_entropy(p_marginal_truth: np.ndarray, p_marginal_alg: np.ndarray, idx_truth: np.ndarray,
idx_alg: np.ndarray, evaluation: Evaluation, is_subgraph: bool = False) -> Evaluation:
"""Calculates the entropy of the truth and algorithm partitions.
Parameters
---------
p_marginal_truth : np.ndarray (float)
the marginal probabilities of the truth partition
p_marginal_alg : np.ndarray (float)
the marginal probabilities of the algorithm partition
idx_truth : np.ndarray (int)
the indexes of the non-zero marginal probabilities of the truth partition
idx_alg : np.ndarray (int)
the indexes of the non-zero marginal probabilities of the algorithm partition
is_subgraph : bool
True if evaluation is for a subgraph. Default = False
Returns
------
evaluation : Evaluation
the evaluation object, updated with the entropy metrics
"""
# compute entropy of the non-partition2 and the partition2 version
entropy_truth = -np.sum(p_marginal_truth[idx_truth] * np.log(p_marginal_truth[idx_truth]))
print('Entropy of truth partition: {}'.format(abs(entropy_truth)))
entropy_alg = -np.sum(p_marginal_alg[idx_alg] * np.log(p_marginal_alg[idx_alg]))
print('Entropy of alg. partition: {}'.format(abs(entropy_alg)))
if is_subgraph:
evaluation.subgraph_entropy_truth = entropy_truth
evaluation.subgraph_entropy_algorithm = entropy_alg
else:
evaluation.entropy_truth = entropy_truth
evaluation.entropy_algorithm = entropy_alg
return evaluation
