If Mean Absolute Error (MAE) loss is not differentiable, how can it be used in neural networks? which majorly are trained using back-propagation

I am wondering if MAE is not differentiable how they can be used as a loss function.

  • $\begingroup$ MAE is differentiable. Refer to this link: stats.stackexchange.com/questions/312737/… $\endgroup$ – Vincent Yong Mar 29 at 3:41
  • 1
    $\begingroup$ So basically MAE is not differentiable at 0 and attaining 0 loss is ideal(in which case we can stop optimization). in all other conditions, MAE is differentiable. $\endgroup$ – Angadishop Mar 31 at 3:53
  • $\begingroup$ Yes it is differentiatable at all values of y_pred except when y_pred is equal to y_true. And I would assume there are very few instances where y_pred is exactly equal to y_true. $\endgroup$ – Vincent Yong Mar 31 at 16:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.