How does a 'BEST' approximation do with extrapolation? #mathformachinelearning

We take a best approximation in L2 and look at how far we can trust the approximation.

[noahsegal]

Thank you! Really intresting and think the conversation like your saying is more of model selection. A lot of predictive analytics realy focus on hopefully being able to perform accurate extrapolation so was thinking of how more complex models like this would take.

[NamasenITN]

What if we don't know what is "the" model? What if there is no closed form model?

[thpdespanolo]

You’re just going to ignore the Runge phenomenon