How is unsupervised learning different from supervised learning?

How is unsupervised learning different from supervised learning?

• A. Unsupervised learning finds structure in unlabeled data, while supervised learning predicts labeled targets.
• B. Unsupervised learning uses labeled targets; supervised uses unlabeled data.
• C. Unsupervised learning requires reinforcement signals.
• D. Unsupervised learning always yields explicit labels.

Answer: (A)

The key idea is how the learning signal uses labels. In unsupervised learning, there are no target labels—the model looks for structure, patterns, or compact representations directly from the data, using goals like grouping similar items (clustering), reducing dimensionality, or estimating the data distribution. In supervised learning, the data comes with labels, and the model learns a mapping from inputs to those labeled outputs by minimizing the difference between its predictions and the true targets, so it can predict those labels on new data. That’s why the correct choice fits best: it captures that unsupervised learning finds structure in unlabeled data, while supervised learning predicts labeled targets. The other statements conflict with these setups—for example, suggesting supervised uses unlabeled data, or that unsupervised requires reinforcement signals, or that unsupervised always yields explicit labels.

The key idea is how the learning signal uses labels. In unsupervised learning, there are no target labels—the model looks for structure, patterns, or compact representations directly from the data, using goals like grouping similar items (clustering), reducing dimensionality, or estimating the data distribution. In supervised learning, the data comes with labels, and the model learns a mapping from inputs to those labeled outputs by minimizing the difference between its predictions and the true targets, so it can predict those labels on new data.

That’s why the correct choice fits best: it captures that unsupervised learning finds structure in unlabeled data, while supervised learning predicts labeled targets. The other statements conflict with these setups—for example, suggesting supervised uses unlabeled data, or that unsupervised requires reinforcement signals, or that unsupervised always yields explicit labels.