Why neural networks are stronger in regressions than statistical regression models

[rajanikant-ghate]

Traditionally statisticians had to experiment with order of polynomial to be used for regressing an input variable to estimate the output variable.
In short, is y = mx + c or y = k(x^2) + k0? And so on…

Universal approximation theorem says, that with a combination of hidden layers, there is a possible approximation of any mathematical function.

And with the computational power today, this makes it super easy to predict using Deep learning.