Mutual Information¶

skfeature.utility.mutual_information

Description¶

Mutual Information provides discrete information-theoretic measures such as information gain, conditional entropy, and symmetrical uncertainty, which are used by filter methods like FCBF and CFS.

Usage¶

import numpy as np

from skfeature.utility import mutual_information

f1 = np.random.randint(0, 5, 200)
f2 = np.random.randint(0, 5, 200)
ig = mutual_information.information_gain(f1, f2)

Functions¶

information_gain(f1, f2): compute IG(f1, f2) = H(f1) - H(f1|f2)
conditional_entropy(f1, f2): compute the conditional entropy H(f1|f2)
su_calculation(f1, f2): compute the symmetrical uncertainty between two discrete features