Paraphrasing Kevin P. Murphy, machine learning is an umbrella term for a collection of methods to automatically learn patterns in data, and then use what we learn to predict future data or to take decisions under uncertainty. ML and statistics are really intertwined subjects, and the connection becomes clear if you took a probabilistic perspective, as Kevin Murphy does in his great book Machine learning: A probabilistic perspective. While these domains are deeply connected at a conceptual and mathematical level, the jargon could make the connection opaque. So, let me bring the ML vocabulary to the problem in this chapter. Using the ML terminology, we say a regression problem is an example of supervised learning. Under the machine learning framework, we have a regression problem when we want to learn a mapping from to , with being a continuous variable.
A machine learner usually talks about features instead of variables. We say that the learning process is supervised because we know the values of the pairs; in some sense, we know the correct answer, and all the remaining questions are about how to generalize these observations (or this dataset) to any possible future observation, that is, to a situation when we know but not .