A label is the target category or value associated with an example in supervised learning. In regression, the label is a real-number output, also called the dependent variable, that a model learns to predict from one or more independent variables. section 16.4.1
Predicting a car’s market price
A dataset can pair each car’s year, engine, and mileage with its market price. The car characteristics are inputs; the market price is the label. Linear regression learns a function that relates those inputs to the price, allowing a new car’s price to be estimated. section 16.4.1
Use the label to distinguish the prediction target from the data used to make the prediction:
- Independent variables describe the example and provide the model’s inputs. section 16.4.1
- The dependent variable is the output associated with those inputs; in the regression example, it is a real number. section 16.4.1
- A model approximates the relation between inputs and labels rather than simply storing the examples. section 16.4.1
