Skip to content


Median Absolute Error#

Median Absolute Error (MAD) is a robust measure of error, similar to MAE, that ignores highly erroneous predictions. Since MAD is a robust statistic, it works well even when used to measure non-normal distributions.

\[ {\displaystyle \operatorname {MAD} = \operatorname {median} (|Y_{i}-{\tilde {Y}}|)} \]


In order to maintain the convention of maximizing validation scores, this metric outputs the negative of the original score.

Estimator Compatibility: Regressor

Score Range: -∞ to 0


This metric does not have any parameters.


use Rubix\ML\CrossValidation\Metrics\MedianAbsoluteError;

$metric = new MedianAbsoluteError();