That is a lot of text for such a simple algorithm, the regular implementation: 1. K observations are chosen randomly from the dataset as initial means. 2. K c

That is a lot of text for such a simple algorithm, the regular implementation: 1. K observations are chosen randomly from the dataset as initial means. 2. K clusters are created by associating every observation with its nearest mean. 3. The centroid of each of the K clusters becomes the new mean. Steps 2 and 3 are repeated until we reach convergence. Very simple cases are badly clustered with this algorithm: https://cdn-images-1.medium.com/max/1200/1*BGSo8t0ktVLDn4ChvH23kA.png https://cdn-images-1.medium.com/max/1200/1*BGSo8t0ktVLDn4Chv... (an example with K=2). However, it is the simplest clustering algorithm thus a good start for people who want to start learning about this field!