Elena Farahbaksh Touli: Challenges in Hierarchical Clustering

Abstract: Hierarchical clustering (HC) is one of the popular methods in the application of data analysis and data science as it is advantageous over flat clustering--a method for clustering a given data with the fixed number of clusters. In general, there are two methods for HC; top-down clustering and agglomerative methods. In this seminar, I will
talk about linkage methods which are agglomerative methods for HC. I will talk about a divisive method for HC as well. The cost function is very simple, but it is NP-hard to minimize. I will also present an approximation algorithm for minimizing the cost function.