The big idea: Ordinary programming means writing the rules yourself.
Machine learning means giving a program examples and letting it work the rules out — which is how problems too messy to write rules for get solved.
What separates the types: What the algorithm is given to learn from.
Labelled examples, unlabelled data, or an environment to experiment in. Answer that and you have named the type.
Free preview
This is the free notes preview
You're reading the free notes. Aimnova Pro unlocks the full study experience — and you can try it with your first topic free to keep:
- FlashcardsLock in vocabulary and key terms with spaced repetition.
- Practice questionsAnswer exam-style questions and get instant AI marking.
- Mock exams & past-paper vaultSit full mocks and see exactly how examiners award marks.
- Personalised study planA daily plan built around your exam date and weak areas.
The same task — telling apples from pears — under each approach. Step through all five.
Interactive diagram
Explore the labelled diagram, charts and maps for this topic in full study mode.
Supervised
- Given labelled examples — each one with its answer
- Learns to predict the label for something new
- Labelling is slow and costly, but progress can be measured
- Medical imaging, sentiment analysis, spam filtering
Unsupervised
- Given data with no labels
- Finds groups or structure the data already contains
- Cannot say what the groups mean — a person must
- Market basket analysis, customer segmentation
Reinforcement
- Given no data — an environment to act in
- Learns which actions earn the most reward over time
- Needs to be allowed to fail a great many times
- Robotics navigation, game playing, control systems
Deep learning
- Many layers, each building on what the last found
- Early layers find edges, later ones whole objects
- Works out its own features — nobody specifies them
- Needs a great deal of data and processing power
- Hard to explain how it reached a given answer
Transfer learning
- Starts from a model already trained on something else
- Only the last part is retrained, on your own data
- Works with hundreds of examples rather than millions
- Far cheaper and quicker than training from scratch
- The new task must be related to the original one
Learn what examiners really want
See exactly what to write to score full marks. Our AI shows you model answers and the key phrases examiners look for.
You have labelled examples — supervised
You have data but no labels — unsupervised
You have neither — reinforcement
You have a little data — transfer
How this is tested — you must choose the approach from WHAT DATA the scenario has. It comes up two ways:
Paper 1 Section A
- Describe the types of machine learning, 3-5 marks
- Name the approach suiting a described problem
- Give a real-world application of one
Paper 1 Section B — case study
- Identify the approach used in the case study
- Justify it from the data that organisation holds
The classic trap: Choosing supervised learning for data that has no labels. Read the scenario for what already exists — if nobody has labelled anything, supervised is not available whatever its advantages.
A streaming service wants to group its users by viewing habits, and separately to predict whether each user will cancel. Describe a suitable approach for each.
Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.