R
Regression — predict a number
The output is a continuous numeric value rather than a category. Also typically supervised, since training data has known numeric outcomes attached.
Example: predicting a house's sale price or tomorrow's temperature — the output can be any number along a continuous range.
C
Clustering — group similar items with no labels
An unsupervised task: the model groups similar data points together without being told in advance what the "correct" groups are.
Example: segmenting customers into behavioral groups based on purchase history, with no pre-existing correct grouping supplied to the model.
G
Generation — produce new content
The model creates new output (text, images, code, audio) rather than labeling or scoring existing input. This is the task type behind LLMs and diffusion models.
Example: ChatGPT writing a new paragraph, or a diffusion model generating a new image from a text prompt — both are generation tasks.
Applied Walkthrough
1
A question describes a model that outputs a single number representing predicted rainfall in millimeters for tomorrow.
2
Ask: is the output a fixed category, or any number along a continuous scale? It's a continuous number.
3
That makes this a Regression task, not classification — classification would require a fixed list of categories like "rainy" or "dry."
4
Contrast: if the same weather model instead output one of three labels — "sunny," "rainy," or "cloudy" — that would switch it to a Classification task instead.
Exam Application
Exams frequently give you a described AI system and ask you to identify its task type, because the task type determines which algorithms, loss functions, and evaluation metrics are even appropriate to use. A continuous numeric output always signals regression; a fixed set of category labels signals classification; no labels at all with a grouping goal signals clustering; and creating brand-new content signals generation.
⚠ Common Trap
The most common trap is confusing regression with classification when a numeric-looking value is actually representing a category (like "1" for spam and "0" for not spam — this is still classification, not regression, because the underlying meaning is a discrete category, not a continuous quantity). Always check whether the number represents a true continuous measurement or is really just a coded label.
✓ Quick Self-Check
1. What kind of output does a classification task produce?
A discrete category label chosen from a fixed set of possible categories.
Tap to reveal / hide
2. What kind of output does a regression task produce?
A continuous numeric value, such as a price or temperature.
Tap to reveal / hide
3. Is clustering a supervised or unsupervised task?
Unsupervised — clustering groups similar items together without any pre-existing correct labels.
Tap to reveal / hide
4. What task type describes an LLM writing a new paragraph of text?
Generation — producing new content rather than labeling or scoring existing input.
Tap to reveal / hide
5. A model outputs "1" for spam and "0" for not spam. Is this regression or classification?
Classification — even though the output looks numeric, it represents a discrete category (spam or not spam), not a true continuous quantity.
Tap to reveal / hide