Can we interpret this phenomenon simply as an overfitting problem? It seems really similar.
I don't see the relationship between this and overfitting... For me it's more like Gibbs phenomenon when we try to recover the square wave with high frequency components.
@SnackMixer I think you're right. The fact that we get worse approximation with higher degree polynomials seems similar to overfitting.
bit confused why we cannot interpolate these high-degree polys to approximate?
Is this the same issue we saw with using Bernstein polynomials of degree >3 for Bezier curves?
I think it is.