ML types
ML is a subset of computer science, data science, and artificial intelligence (AI) that enables systems to learn and improve from data without additional programming.
Instead of using explicit instructions for performance optimisation, ML models rely on algorithms and statistical models that deploy tasks based on data patterns and inferences. In other words, ML leverages input data to predict outputs, continuously updating outputs as new data becomes available.
On retail websites, for instance, machine learning algorithms influence consumer buying decisions by making recommendations based on purchase history. Many retailers’ e-commerce platforms, including those of Woolworths, Amazon, Google, Meta and Netflix, rely on artificial neural networks (ANNs) to deliver personalised recommendations. And retailers frequently leverage data from chatbots and virtual assistants, in concert with ML and natural language processing (NLP) technology, to automate users’ shopping experiences.
Syllabus alignment
This lesson supports the NSW Software Engineering Stage 6 syllabus:
-
Software automation / Algorithms in machine learning
- Explore models of training ML, including supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning.
Why do different learning paradigms exist?
Machine learning systems confront many kinds of problems: some come with neat answer keys, some arrive as piles of unlabeled data, and others require acting in a world that pushes back. Each paradigm—supervised, unsupervised, and reinforcement learning—evolved to match these different information landscapes. Choosing wisely matters because mismatched paradigms can waste data, amplify bias, or learn the wrong objective entirely.
Another reason for multiple paradigms is the practical balance between data availability and feedback quality. Supervised learning thrives when labeled datasets are plentiful; unsupervised learning shines when labels are scarce but structure exists; reinforcement learning tolerates delayed feedback by converting experiences into rewards. Understanding these trade-offs helps students diagnose which paradigm will unlock progress on a given problem.
01 Supervised learning
Definition: Supervised learning trains models on labeled examples, where each input is matched to the correct output. The model learns to map inputs to outputs by minimizing errors on known answers.
Real-world scenario: Email providers use supervised learning to filter spam. Engineers feed in thousands of emails tagged as “spam” or “not spam,” and the model learns to classify new messages reliably.
Analogy for students: Imagine studying with a teacher who grades every practice quiz instantly. You try a question, check the answer, adjust your thinking, and gradually improve because you always know whether you were right.
02 Unsupervised learning
Definition: Unsupervised learning explores unlabeled data to uncover patterns, clusters, or latent structures without explicit right-or-wrong answers.
Real-world scenario: Music streaming services group songs with similar audio features to build auto-generated playlists, even when no human has labeled the tracks by mood or style.
Analogy for students: Picture walking into a new school cafeteria and grouping people by the conversations you overhear. No one tells you who belongs together; you infer the cliques by observing how people naturally cluster.
03 Reinforcement learning
Definition: Reinforcement learning trains an agent to make decisions by interacting with an environment and receiving feedback in the form of rewards or penalties.
Real-world scenario: Autonomous warehouses use reinforcement learning robots to navigate aisles efficiently, rewarding them for fast, collision-free deliveries and penalizing mistakes.
Analogy for students: Think of playing a video game where you learn which actions earn points or cause you to lose lives. Through repeated play, you refine your strategy to maximize your score.
Comparative summary
| Paradigm | Typical Data | Learning Signal | Strengths | Challenges | Example Use Case |
|---|---|---|---|---|---|
| Supervised | Labeled pairs of inputs and outputs | Direct error feedback on predictions | High accuracy on well-defined tasks | Requires large, clean labeled datasets | Diagnosing diseases from annotated medical images |
| Unsupervised | Unlabeled data | Implicit patterns discovered in the data | Finds hidden structure and reduces dimensionality | Hard to evaluate success without labels | Customer segmentation for marketing campaigns |
| Reinforcement | Experience from interactions | Rewards and penalties over time | Learns adaptive, sequential decision policies | Requires careful reward design and extensive exploration | Training game-playing AIs like AlphaGo |
Reflective prompt
Discussion question: Which paradigm would you choose to build a study companion app that suggests practice problems, and why? Consider the data you would have, how feedback would work, and what success looks like.
Knowledge Check
Now that you’ve read about the four types of machine learning, complete the table below to summarise your understanding. Copy and paste the table below into your digital notebook and complete each cell.
| Aspect | Supervised Learning | Unsupervised Learning | Semi-Supervised Learning | Reinforcement Learning |
|---|---|---|---|---|
| Data type Used | ||||
| Learning style | ||||
| Example application | ||||
| Best used when | ||||
| Example in practice |
Hints
- Supervised: Think “learning with a teacher” - you have the answers
- Unsupervised: Think “discovering patterns” - no answers provided
- Semi-supervised: Think “limited help” - few answers, lots of unlabeled data
- Reinforcement: Think “trial and error” - learning through rewards and penalties