JSD - Jensen-Shannon Divergence

• Best possible score is 0.0 (identical), smaller value is better . Range = [0, +inf)

• Link: https://machinelearningmastery.com/divergence-between-probability-distributions/

Example to use JSD metric:

from numpy import array
from permetrics.regression import RegressionMetric

## For 1-D array
y_true = array([3, -0.5, 2, 7])
y_pred = array([2.5, 0.0, 2, 8])

evaluator = RegressionMetric(y_true, y_pred, decimal=5)
print(evaluator.jensen_shannon_divergence())

## For > 1-D array
y_true = array([[0.5, 1], [-1, 1], [7, -6], [1, 2], [2.1, 2.2], [3.4, 5.5]])
y_pred = array([[0, 2], [-1, 2], [8, -5], [1.1, 1.9], [2.0, 2.3], [3.0, 4.2]])

evaluator = RegressionMetric(y_true, y_pred, decimal=5)
print(evaluator.JSD(multi_output="raw_values"))