The British historian Arnold Toynbee once said, "no tool is omnicompetent". In this chapter, we used three tools for clustering. Each of the three algorithms we discussed here approaches the problem from a different angle. The K-means clustering algorithm tries to find points that summarize the clusters and the centroids and builds its clusters around them. The agglomerative clustering approach is more of a bottom-up approach, while the DBSCAN clustering algorithm introduces new concepts such as core points and density. This chapter is the first of three chapters to deal with unsupervised learning problems. The lack of labels here forced us to learn about newer evaluation metrics, such as the adjusted rand index and the silhouette score.
In the next chapter, we are going to deal with our second unsupervised learning problem: anomaly detection. Luckily, the concepts discussed here, as well as the ones fromChapter 5, Image Processing with Nearest Neighbors...