K-Means

Properties2

Type	Concept
Note created	Mar 20, 2025

K-Means is a clustering algorithm that generates a set of $K$ clusters (defined as an input to the algorithm) based on a set of input dimensions in an unsupervised fashion.

The steps of the algorithm consist of:

1. Define $K$ centroids using an initialization method (can be at random).
2. Assign each input sample to the closest centroid.
3. Update the centroid using the mean of each dimensions of all the assigned points.
4. Repeat until convergence (the updates are no longer significant) or a max number of iterations is reached.