k-Means

Describing the algorithm breifly:

1. choose k (hyperparameter) : the number of clusters/classes
2. randomly put k centroids in the hyperspace
3. compute distance (euclidean usually) of each sample to each cetroid
4. assign closest centroid to each example
5. for each centroid group, compute the new centroid
6. repeat from 3 until convergence, i.e. centroids are stable and don't change anymore

Do note that different initializations can result in different final results for the algorithm. Choosing the number of clusters (k) is more about making an educated guess than a science although some heuristics do exist for the same.