[jpfr]

Neural networks are all "nonlinear function approximation". Last I checked, the state-of-the-art in all the fields you mention is some variation of deep neural network.

Not to be despicable. On the contrary, function approximation is a very rich and deep topic.