Supervised Learning
Learning with an Answer Key
6 min read
Learning with an Answer Key
Studying for an exam where every practice question has an answer key — you know immediately if you got it right.
In school, your teacher gives you 1000 maths problems with solutions. You practice, check your answers, learn from mistakes, and get better. Supervised Learning is the same: feed the AI thousands of examples with correct labels (this photo = cat, this email = spam), and it learns to label new examples correctly.
In Plain English
Supervised Learning is training an AI using examples where the correct answer is already provided. The AI learns by comparing its guesses to the correct answers and adjusting until it gets them right.
The Technical Picture
In supervised learning, a model is trained on labelled data pairs (input, label). The model minimises a loss function measuring the difference between predictions and ground truth labels across the training set, using gradient descent to update parameters.
Real-World Examples
- Email spam filters trained on millions of labelled spam/not-spam emails
- Medical image AI trained on X-rays labelled by doctors
- Sentiment analysis trained on customer reviews labelled as positive/negative
Supervised Learning = AI learning from labelled examples, like a student with an answer key.