colsplit_mutual_information#

tangles.util.entropy.colsplit_mutual_information(data: ndarray, partitions: ndarray, combined_entropy='joint_entropy') → ndarray#

Mutual information in the two sides of partitions. The partitions are of the column vectors of a data matrix.

Parameters#

datanp.ndarray: The data.
partitionsnp.ndarray: A matrix with partition-indicator-vectors in its columns. Each column represents a partition. It has shape \((k, l)\), where \(k\) is the number of columns in data and \(l\) is the number of partitions. We assume that the partition-indicator-vectors split the sides by negative vs. non-positive entries.
combined_entropy{‘joint_entropy’, ‘max_entropy’}: How to calculate the entropy of both sides of the partition together. Either ‘joint_entropy’ (of the entire data) or ‘max_entropy’ (of both sides).