Learning from unlabeled data to discover structure such as clusters or latent factors; includes methods like autoencoders and k‑means.