tf.keras.ops.nanvar

Compute the variance along the specified axes, ignoring NaNs.

tf.keras.ops.nanvar(
    x, axis=None, keepdims=False
)

x Input tensor.
axis Axis or axes along which the variance is computed. The default is to compute the variance of the flattened tensor.
keepdims If True, the axes which are reduced are left in the result as dimensions with size one. Defaults to False.

Output tensor containing the variance ignoring NaNs.

Examples:

import numpy as np
from keras import ops
x = np.array([[1.0, np.nan, 3.0],
              [np.nan, 2.0, 1.0]])
 
ops.nanvar(x)
0.6875
 
ops.nanvar(x, axis=1)
array([1.  , 0.25])
 
ops.nanvar(x, axis=1, keepdims=True)
array([[1.  ],
       [0.25]])