Supervised vs unsupervised vs reinforcement learning, with Indian examples

The three main ways machines learn, explained with everyday Indian examples: predicting loan defaults, segmenting shoppers, and learning by trial and error. Plus where self-supervised learning fits.

Supervised learning trains a model on examples with known answers, such as past loans labelled repaid or defaulted. Unsupervised learning finds structure in data without answers, such as grouping similar shoppers. Reinforcement learning trains an agent by trial and error with rewards, such as optimising a delivery route. Most real projects use supervised learning first.

These three categories are the first thing every machine learning course teaches, and the first thing interviewers ask about. Here they are with examples you will recognise, the algorithms behind them and how to choose.

What is supervised learning?

In supervised learning, every training example has an input and the correct output, called a label. The model learns the mapping and then predicts labels for new inputs. It comes in two main forms: classification, where the answer is a category, and regression, where it is a number.

What is unsupervised learning?

Unsupervised learning works with data that has no labels and looks for patterns: groups, unusual points or simpler representations.

What is reinforcement learning?

In reinforcement learning, an agent takes actions in an environment and receives rewards or penalties. Over many attempts it learns a strategy that maximises long-term reward. It suits sequential decisions: game playing, robotics, resource allocation and, importantly, the RLHF step used to align language models. The standard textbook, Reinforcement Learning: An Introduction by Sutton and Barto, is freely available online.

Where does self-supervised learning fit?

Self-supervised learning creates labels from the data itself. Large language models are trained this way: the “label” for each position is simply the next token in the text. It sits between supervised and unsupervised learning and is why models can learn from enormous amounts of unlabelled text. Our explainer on transformers shows how that training works.

How do you choose the right approach?

What does a supervised learning project look like end to end?

Take loan default prediction for a small finance company. You gather past loans with outcomes, clean the data, engineer features such as repayment history and income ratios, split the data by time so the model is tested on later loans, train a baseline and a stronger model, and evaluate with metrics that reflect cost, such as how many defaulters you catch for each good customer wrongly flagged. Finally, you explain the model’s decisions, because lending decisions must be justifiable.

Is semi-supervised learning useful?

Yes, when labels are expensive. Semi-supervised methods train on a small labelled set plus a larger unlabelled one, for example a few hundred labelled support tickets and thousands of unlabelled ones. Combined with pretrained models, this often gets surprisingly good results with limited labelling effort.

What mistakes do beginners make?

Our guides to overfitting and the bias-variance trade-off and data science portfolio projects cover evaluation in more depth.

How do these approaches show up in Indian industry?

Banks and lenders use supervised models for credit risk and fraud; retailers and food platforms use clustering for segmentation and recommendations; logistics companies optimise routes and warehouse operations, sometimes with reinforcement learning ideas; and language technology for Indian languages relies heavily on self-supervised pretraining. Knowing which category a problem belongs to is the first step in solving it well.

Where can you learn these in order?

Program Zero’s data science and classical machine learning phase covers supervised and unsupervised learning with real projects, after the data science foundations, and later applies reinforcement learning ideas in LLM alignment.

Program Zero runs live online, so you can learn this from anywhere in India: metros such as Bengaluru, Pune, Hyderabad and Noida, growing cities such as Lucknow, Kanpur, Visakhapatnam and Madurai, and smaller towns such as Siliguri, Jamshedpur, Durgapur and Cuttack. Every learner joins the same evening classes, at the same fee.

Frequently asked questions

Related reading