Let there be a “training set” such that is an “input” value and has a corresponding “output” or “target” value and a single pair is called as a “training example” Let be the space of input values and be the space of output values. Formally, the goal of supervised learning is to find a function, historically referred to as a hypothesis, .

We shall differentiate between problems where is continuous and discrete by referring to them as regression and classification problems respectively.