Which statement specifically applies to K-means clustering and not to hierarchical clustering?

K-means clustering specifically requires an initial set of centroids to start the clustering process, which means it must be initialized. This initialization can significantly influence the final clusters formed because K-means relies on these initial centroids to partition the data into groups. During the algorithm's iterations, points are assigned to the nearest centroid, and the centroid positions are then recalculated based on the mean of the assigned points. Hence, if the initial centroids are poorly chosen, the algorithm can converge to suboptimal cluster configurations.

In contrast, hierarchical clustering builds a tree-like structure called a dendrogram without requiring an initial assignment of data points to clusters. Instead, it systematically merges or splits clusters based on a defined distance metric, progressively forming larger clusters. Therefore, the dependence on an initial assignment is a key distinguishing feature of K-means clustering that does not apply to hierarchical clustering.