Chapter 2

1. The Manhattan distance is the same as the Minkowski distance with p=1; hence, we expect to observe a longer distance.
2. No; the convergence speed is primarily influenced by the initial position of the centroids.
3. Yes; k-means is designed to work with convex clusters, and its performances are poor with concave ones.
4. It means that all clusters (except for a negligible percentage of samples), respectively, only contain samples belonging to the same class (that is, with the same true labels).
5. It indicates a moderate/strong negative discrepancy between the true label distribution and the assignments. Such a value is a clear negative condition that cannot be accepted, because the vast majority of the samples have been assigned to the wrong clusters.
6. No, because the adjusted Rand score is based on the ground truth (that is, the expected number of clusters is fixed).
7. If all of...