In part 3, we apply what you’ve learned about mathematical functions, vectors, and calculus to implement some machine learning algorithms. We hear a lot of hype around machine learning, so it’s worth being precise about what it actually is. Machine learning is part of the field of artificial intelligence, or AI, which studies how to write computer programs to accomplish tasks intelligently. If you’ve ever played a video game against a computer adversary, you’ve interacted with an artificial intelligence. Such an adversary is programmed (usually) with a set of rules which help it destroy you, outmaneuver you, or otherwise defeat you.
For an algorithm to be classified as machine learning, it must not only operate autonomously and intelligently, but it must learn from experience. That means that the more data it receives, the better it performs at the task at hand. The next three chapters focus on a specific kind of machine learning called supervised learning. When we write supervised learning algorithms, we give them training data sets with pairs of inputs and corresponding outputs, and the algorithms should then be able to look at new inputs and come up with correct outputs on their own. In this sense, the result of training a machine learning algorithm is a new mathematical function that can effectively map some kind of input data to some kind of decision as an output.