Huber/Smooth L1/Smooth MAE

• It is less sensitive to outliers than the MSE and in some cases prevents explodingarchitecture
• Fast-RCNN

if $\left(\Vert y-\hat{y}\Vert <1.0\right)>1$

$\frac{1}{\mathrm{length}\left(y\right)}\cdot \mathrm{sum}\left(0.5\cdot {\left(y-\hat{y}\right)}^2\right)$

else

$\frac{1}{\mathrm{length}\left(y\right)}\cdot \mathrm{sum}\left(\Vert y-\hat{y}\Vert -0.5\right)$

…