Learning Two Uncorrelated Models

Data Science
15

Here is a version for Tensorflow based on the code of @jrb which can be used for a custom loss:

import tensorflow as tf
import tensorflow_probability as tfp

def correlation(x,y):
    mx = tf.math.reduce_mean(x)
    my = tf.math.reduce_mean(y)
    xm, ym = x - mx, y - my
    r_num = tf.math.reduce_mean(xm * ym)
    r_den = tfp.stats.stddev(xm) * tfp.stats.stddev(ym)
    return r_num / (r_den + tf.constant(0.00001, dtype=x.dtype))
2 Likes