Mean error summarizes how far modeled values typically fall from reference values across an evaluation set. For a polynomial sine approximation, the mean error exceeds 0.003, even though interpolation points have zero error. section 7.1.2
The average alone does not show where errors occur: they are unevenly distributed, and missed derivative behavior can make a circle generated from the model visibly imperfect. section 7.1.2
Comparing sine approximations
Polynomial interpolation produces mean error above 0.003, whereas a Maclaurin-series model reaches relative error as low as 0.000018 and produces a much smoother circle. section 7.1.2
