Use of mean absolute error has been rising. It is not as amenable to closed-form analysis, but modern numerical techniques should be able to cope. MAE is much more intuitive.
Hi Ryan, welcome to Xeek!
We considered both RSME and MAE, but we chose RSME because we wanted a scoring method where larger errors are penalized higher than smaller errors. If all errors (large and small) have the same weight we would use MAE.
The example in the documentation doesn’t show the difference between the X_pred_row and X_actual_row, Y_pred_row and Y_actual_row, Z_pred_row and Z_actual_row as squared before the differences are summed and the square root is taken.
I’m assuming the error should be the sum of the square root of the sum of these differences squared for all points.
Do you know if this is correct?