Description acceptable. The optimal choice for this quantity maximizes the Caliński–Harabasz (“kah-leen-skee hahr-ah-bash”) index. Chinese restaurant processes increment this quantity at each step with a probability that depends on the alpha parameter of the Dirichlet (“deer-ih-CLAY”) distribution. A good value for this quantity gives a high average silhouette score and evenly proportioned silhouette plots. Agglomerative approaches to building a hierarchy decrease this quantity at each step, as shown on dendrograms. This quantity can be selected by finding an “elbow” on plots of WCSS versus it. This quantity is selected as an input to, and lends its name to, an algorithm whose name ends in “plus plus,” which outperforms Lloyd’s algorithm and iteratively updates Voronoi cell centroids. For 10 points, name this quantity symbolized “k” in an unsupervised learning algorithm named for choosing “k means.” ■END■
ANSWER: number of clusters [accept number of partitions; accept number of tables after “Chinese” is read; accept number of centroids until “centroids” is read; accept number of means until “means” is read; accept descriptions of how many clusters or how many partitions; prompt on k until read] (WCSS is the within-cluster sum of squares. Lloyd’s algorithm is improved upon by k-means++.)
<Other Science>
= Average correct buzz position