Unsupervised Learning

In this kind of learning, we do not have labeled training data.

The most common unsupervised problem is clustering. There are also other unsupervised problems such as dimensionality reduction.

Some applications of clustering include:

• color quantization (16 million colors to 256 colors)

• directed marketing (different ads for different groups of people)

• pre-processing stage in supervised learning (in certain cases where it is clear that the examples are coming from well-defined groups)