PPT Slide
In the context of Bayessian regularization, the assumption
behind the quadratic terms, it is that the errors:
had a Gaussian (normal) distribution.
Because of discontinuities in the data and noise,
errD and errR are “corrupted” by outliers: