John I try to get you some interim computations to visualize how the minimization of error is handled by ML algorithms e.g. SVR.

Problem is that there are very simple interpolation functions that reduce the error 0, they are very wavy and some might call them over-fitting.

So each minimization algorithm has cooked up a constrained to avoid that over-fitting.

Try to cook something up with read data next couple of days

Dara