Runge's Phenomenon and the Perils of Polynomial Curve Fitting

Here is more about curve fitting and Runge’s phenomenon. The pathological behaviour of polynomials fit to data can be minimised by using Chebyshev nodes as the points which are fit.


In contemporary machine learning, the two most frequently used tools are regularization and Bayesian priors.

I like the principled nature of priors, but regularization is often effective and computationally simpler.

1 Like