zscore¶

odl.util.numerics.zscore(arr)[source]¶

Return arr normalized with mean 0 and unit variance.

If the input has 0 variance, the result will also have 0 variance.

Parameters:

Returns:

Examples

Compute the z score for a small array:

>>> result = zscore([1, 0])
>>> result
array([ 1., -1.])
>>> np.mean(result)
0.0
>>> np.std(result)
1.0

Does not re-scale in case the input is constant (has 0 variance):

>>> zscore([1, 1])
array([ 0., 0.])