Overfitting

Description

The sample data is what we use for fitting the curves and test data is what we use to evaluate how good our fit is. The input error

is the avarage squared error to to the sample data (so

is the i'th data point in the sample data). Similarly the output error is given by

where

is the i'th data point in the test data. We want

as small as possible. As we see

generally performes much better that

You can freely investigate the definitions of the components in the left menu. For instance try to click on

and

and 'sampledata'. You will be surprised to se how easy lagrange interpolation is in GeoGebra!