Supervised learning trains a model using examples that pair inputs with correct outputs. This lesson introduces the basic idea, common task types, and the practical steps used to build and evaluate a supervised learning system.
In supervised learning, an algorithm uses labeled examples to learn a function that maps an input to an output. Each training example contains an input, such as an image, and its corresponding correct label, such as “cat.” The label acts like a teacher’s answer key, guiding the algorithm during training.
All flashcards
What is supervised learning?
A machine-learning approach in which an algorithm learns to map inputs to desired outputs from labeled input-output examples.
What does generalization mean?
The ability of a trained model to make useful predictions for new examples that were not used during training.
How do classification and regression differ?
Classification predicts a category or class; regression predicts a numerical value.
What is a feature vector?
A representation of an input as a collection of features that describe properties relevant to predicting its output.
What roles do training, validation, and test sets serve?
Training data fit the model, validation data support tuning and model choices, and test data provide a final evaluation on held-out examples.
Quiz questions
What makes a training example labeled in supervised learning?Answer: It includes an input together with its correct or desired output.
Which supervised task predicts a category such as spam or not spam?Answer: Classification.
What is the main purpose of a validation set?Answer: To help tune parameters and make model choices before final testing.
A model predicts a numerical sale price for each house. Which task type is this?Answer: Regression.
Why should final performance be measured on a test set separate from training?Answer: Because performance on separate data better estimates how the model will perform on unseen examples.