WikiMath

In the following, we present various well-known clustering techniques, from the most classical to the least frequent. You are asked to discover these various techniques, then to carry out clustering studies, evaluating them on data sets from sklearn (Iris, etc., see link).

• Classic clustering:
  • K-Means
    • Mini Batch K-Means
    • K-Medoids
    • K-Modes
  • Hierarchical clustering
    • Agglomerative
    • Divisive: Unlike hierarchical clustering, the divisive starts with a single cluster encompassing all points, then iterates divisions until only clusters at one point are obtained.
  • Mean-shift
  • Nearest neighbors
  • Birch
  • DBSCAN
    • HDBSCAN*
  • Gaussian Mixture Model
  • Affinity propagation
  • MCL Markov clustering
  • Spectral clustering
• Clustering under constraints:
  • Active semi-supervised clustering algorithms
  • COP-Kmeans: COP (COnstrained Pairwise) K-Means is a constrained clustering technique, where it is specified that this and that individual must be linked, when that and that other must not.
  • MinSizeKmeans : forces a minimum size for clusters.