how to compute norm.ppf() in tensorflow/Keras?
Question:
I want to use inverse of cumulative distribution function (cdf) which can be done using norm.ppf() from scipy https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.norm.html#scipy.stats.norm
inside a layer of a tf/keras model architecture.
Answers:
As per documentation, tfp.distribution.Normal does have a method for calculating ppf (percent point function). It’s called quantile:
scipy.stats.norm(loc=0, scale=1).ppf(0.95)
Output:
1.6448536269514722
Tensorflow:
tfp.distributions.Normal(loc=0, scale=1).quantile(0.95)
Output:
<tf.Tensor: shape=(), dtype=float32, numpy=1.6448536>
I want to use inverse of cumulative distribution function (cdf) which can be done using norm.ppf() from scipy https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.norm.html#scipy.stats.norm
inside a layer of a tf/keras model architecture.
As per documentation, tfp.distribution.Normal does have a method for calculating ppf (percent point function). It’s called quantile:
scipy.stats.norm(loc=0, scale=1).ppf(0.95)
Output:
1.6448536269514722
Tensorflow:
tfp.distributions.Normal(loc=0, scale=1).quantile(0.95)
Output:
<tf.Tensor: shape=(), dtype=float32, numpy=1.6448536>