Two Kinds of Questions

Supervised learning always starts from labeled examples, but the shape of the label itself splits it into two families: classification, where the output is one of a fixed set of categories, and regression, where the output is a continuous number.

Think of email spam detection: every email gets sorted into one of two categories, 'spam' or 'not spam.' The set of possible answers is fixed and known ahead of time — the model is only ever choosing among those categories, never inventing a new one.

Now think of predicting a house's price: the output isn't one of a handful of pre-set choices, it's a number that could land almost anywhere — $210,000, $647,500, $1.2 million. That's regression: the target is a continuous number, not a category.

A quick way to tell them apart: if you can list every possible right answer in advance, it's classification. If the right answer could be any value along a range, and listing them all doesn't make sense, it's regression. Predicting a customer's risk tier (low, medium, high) is classification; predicting their exact credit score is regression.

Which of these is a classification problem?

A supervised learning task where the output is a continuous number, like a price or a temperature, is called ____.

In a classification problem, the model can output a brand-new category that never appeared in the training examples.