Unsupervised learning is a category of machine learning in which the model is trained on unlabeled data. Unlike supervised learning, there are no predefined outputs or targets. Instead, the algorithm tries to find hidden patterns or structure in the data on its own.
Common tasks in unsupervised learning include clustering (grouping similar data points, as in k-means) and dimensionality reduction (simplifying data while preserving important structures, as in PCA or autoencoders). These techniques are useful for data exploration, anomaly detection, data compression, and pre-processing.
Unsupervised learning is powerful when labeled data is unavailable or expensive to obtain. Since the learning is not guided by explicit answers, it can be more challenging to evaluate model performance. Nevertheless, it’s a crucial tool for discovering relationships in data and is widely used in customer segmentation, recommendation systems, and natural language processing.
In reinforcement learning, unsupervised techniques can be used for representation learning — helping agents learn useful features from raw inputs without reward signals — or for intrinsic motivation, encouraging exploration through curiosity.