Unsupervised

This is where a learning algorithm is presented with data without any other information (label) the goal is then to find any patterns in the data.

Google did this For example to make a face detector, learning by only using a large dataset of unlabeled images, 10 million 200x200 pixel images downloaded from the Internet. Google also found cat faces in random web pictures.

As in the face learning the use of "Big data" MANY examples are often used in unsupervised learning.

In pure Unsupervised learning Input data 'X' is not labelled at all, but there are also other ways of adding weighting 'w' in order for the algorithms to work more akin to the way our own brains neural networks learn.