Unsupervised learning finds patterns in unlabeled data

In unsupervised machine learning, an algorithm gets fed a dataset without any answers. The machine is supposed to find structure in the input itself.

Depending on the approach, the algorithm could organise extracted patterns in the following ways:

  • Clustering: Placing objects that are more similar (in some sense) than others in groups
  • Anomaly detection: identifying rare items that vary from the majority of the data
  • Neural Networks: connecting units in a way that models the connection of neurons in a biological brian
  • Learning latent variable methods (?)