Numerical approximation and error
Absolute error is the absolute difference between approximation and reference; relative error additionally requires a nonzero reference and divides by its absolute value.
atlas-numerical-approximation-error · canonical_mko