TL;DRSupervised learning trains on labeled examples — input paired with the correct answer.
You give the model inputs and their known outputs (labels): images tagged "cat"/"dog", emails marked spam/not. It learns to predict the label for new inputs. It's the most common, most reliable paradigm — but it needs labeled data, which is expensive to produce. Classification (discrete labels) and regression (continuous values) are its two flavors.
Key points
Trains on (input, correct-label) pairs
Predicts labels for new inputs
Needs labeled data (costly)
Classification + regression
Common mistakes
Underestimating labeling cost/effort
Label leakage inflating accuracy
Assuming labels are always correct
Try it: Give one classification and one regression example.
Example code
<!doctype html><html><head><meta charset="utf-8"></head>
<body style="background:#06040d;color:#e6e0ff;font-family:monospace;padding:20px"><pre>labeled data: (image, "cat"), (image, "dog") ...
→ model predicts label for a new image
classification: cat/dog · regression: house price</pre></body></html>