Unsupervised learning
Machine learning from unlabeled data: the model finds structure such as clusters or patterns without being told the right answers.
Unsupervised learning finds structure in unlabeled data: no target outputs are given, so the model groups or compresses examples by similarity alone.
It is a core strand of machine learning, the counterpart to supervised learning on labeled pairs and to reward-driven reinforcement learning. Classic methods include clustering (grouping similar points, as in k-means) and dimensionality reduction, where a model organizes raw examples on its own. GANs (2014) learn to generate synthetic data by pitting two networks against each other without labelled targets. The self-supervised pretraining behind modern deep learning is closely related: it learns from raw text with no human labels before any fine-tuning.