Hands-On ROS for Robotics Programming

Clustering

Clustering is a scenario where we have many items and we want to group them by similarity. In this case, items are unlabeled and we ask the algorithm to do two things:

• Make a group of similar items.
• Label these groups so that new items are both classified and labeled by the algorithm.

As an example, think of a collection of texts about many topics and you wish the algorithm to group similar texts and identify the main topic of each group, that is, label them: history, science, literature, philosophy, and so on. One of the classical algorithms for this scenario is the nearest neighbor method, where you define a metric, calculate it for each pair of items, and group together those pairs that are close enough (based on the defined ...