The work by Hornik et al (http://www.cs.cmu.edu/~bhiksha/courses/deeplearning/Fall.2016/notes/Sonia_Hornik.pdf) proved the following:
"multilayer feedforward networks with as few as one hidden layer are indeed capable of universal approximation in a very precise and satisfactory sense."
In this recipe, we will show you how to use MLP for function approximation; specifically, we will be predicting Boston house prices. We are already familiar with the dataset; in Chapter 2, Regression, we used regression techniques for the house price prediction, now we will do the same using MLPs.