3.5. Summary
29
tor of a Laplacian [91]. Since then, spectral clustering has been investigated and
extended by many researchers. Applications of the spectral approach were made
popular by the studies of Shi, Malik [84] and Ng et al. [71]. They developed two
versions of normalized spectral clustering algorithms. They also interpreted the idea
of clustering in two different ways, minimum cut and random walk. The non-linear
counterpart, p-Laplacian clustering algorithm was developed by B¨ uhler and Hein
[10]. Subsequently, many extended spectral clustering approaches have been presented, including spectral co-clustering [23], image segmentation based on spectral
clustering [33], detecting the community structures in networks [69] and improved
spectral clustering algorithms [54, 55]. An overview over the properties and interpretation of Laplacian clustering can be found in von Luxburg [101] and an overview
focused more directly on social networks in Seary and Richards [81].
29
tor of a Laplacian [91]. Since then, spectral clustering has been investigated and
extended by many researchers. Applications of the spectral approach were made
popular by the studies of Shi, Malik [84] and Ng et al. [71]. They developed two
versions of normalized spectral clustering algorithms. They also interpreted the idea
of clustering in two different ways, minimum cut and random walk. The non-linear
counterpart, p-Laplacian clustering algorithm was developed by B¨ uhler and Hein
[10]. Subsequently, many extended spectral clustering approaches have been presented, including spectral co-clustering [23], image segmentation based on spectral
clustering [33], detecting the community structures in networks [69] and improved
spectral clustering algorithms [54, 55]. An overview over the properties and interpretation of Laplacian clustering can be found in von Luxburg [101] and an overview
focused more directly on social networks in Seary and Richards [81].