Addressing Luigi's comments: adding explanations on universal approx.thm., stochastic grad.desc. etc.