University of Illinois at Urbana-ChampaignUrbana, ILjliu64@illinois.edu
University of Illinois at Urbana-ChampaignUrbana, ILhanj@illinois.edu
In this chapter, we introduce the family of spectral clustering algorithms which have seen increasing popularity over the past few years. Starting with the seminal works in  and , a large number of papers has been published along this line of work. As opposed to “traditional clustering algorithms” such as k-means and generative mixture models which always result in clusters with convex geometric shape, spectral clustering can solve problems in much more complex scenarios, such as intertwined spirals, or other arbitrary nonlinear ...