Hands-On Unsupervised Learning with Python

Chapter 4

1. In the agglomerative approach, the algorithm starts from each sample, considered as a cluster, and proceeds to merge the sub-clusters until a single cluster has been defined. In the divisive approach, the algorithm starts from a single cluster containing all of the samples, and proceeds by splitting it until every sample makes up a cluster.

2. The closest points are (0, 0) and (0, 1), so the single linkage is L_s(a, b) = 1. The farthest points are (-1, -1) and (1, 1), so the complete linkage is L_c(a, b) = 2√2.
3. No; a dendrogram is a tree representation of a hierarchical clustering procedure, given a metric and a linkage.
4. In an agglomerative clustering, the initial part of a dendrogram contains all of the samples as autonomous clusters. ...