Unsupervised Machine Learning: Clustering

There are 3 main categories that make up machine learning as a whole:

• Supervised learning, where your training examples are labeled,

• Unsupervised learning, where your training data is unlabeled,

• Reinforcement learning, where an agent learns in an environment through trial and error—given feedback.

In this issue, we’re going to focus on the middle (and unpopular) option; unsupervised learning.

• In particular, there are 2 main problems that unsupervised learning algorithms solve.

• There is clustering (grouping ungrouped data) and dimensionality reduction (lowering the number of features/dimensions of the data).

We’ll focus on clustering, as it is the most used and popular form of unsupervised learning.

Clustering deals with finding structure in a collection of unlabeled data.

• If the data were to be labeled, it would be a supervised learning problem.

A loose definition of clustering could be “the process of organizing objects into groups whose members are similar in some way”.

A cluster is therefore a grouping of objects or data points that are more similar to each other in the cluster than to those in other clusters.

Some applications of clustering include market segmentation (dividing potential clients into certain groups), anomaly detection (discovering non-normal data points), and image segmentation (image classification, recognition).

Different clustering algorithms have different ways of sorting data into various clusters. Let’s investigate a few: