lVIULTIWAY CUTS AND SPECTRAL CLUSTERING
نویسنده
چکیده
\eVe look at spectral clustering as optimization. \eVe show that near some special points called perfect, spectral clustering optimizes simultaneously two criteria: a dissimilarity measure that we call the rrmltiway normalized c'ut (lvfNC'I1t) and a cluster coherence measure that we call the gap. The immediate implication from the user's p.o.v is that spectral clustering will optimize any tradeoff between kINOllt and gap which may explain its success in practice. Finally, we propose new methods for selecting K based on the gap and show their superior performance in experiments.
منابع مشابه
A Feature Space View of Spectral Clustering
The transductive SVM is a semi-supervised learning algorithm that searches for a large margin hyperplane in feature space. By withholding the training labels and adding a constraint that favors balanced clusters, it can be turned into a clustering algorithm. The Normalized Cuts clustering algorithm of Shi and Malik, although originally presented as spectral relaxation of a graph cut problem, ca...
متن کاملBeyond Spectral Clustering - Tight Relaxations of Balanced Graph Cuts
Spectral clustering is based on the spectral relaxation of the normalized/ratio graph cut criterion. While the spectral relaxation is known to be loose, it has been shown recently that a non-linear eigenproblem yields a tight relaxation of the Cheeger cut. In this paper, we extend this result considerably by providing a characterization of all balanced graph cuts which allow for a tight relaxat...
متن کاملNormalized cuts clustering with prior knowledge and a pre-clustering stage
Clustering is of interest in cases when data are not labeled enough and a prior training stage is unfeasible. In particular, spectral clustering based on graph partitioning is of interest to solve problems with highly non-linearly separable classes. However, spectral methods, such as the well-known normalized cuts, involve the computation of eigenvectors that is a highly time-consuming task in ...
متن کاملAutomatically finding clusters in normalized cuts
Normalized Cuts is a state-of-the-art spectral method for clustering. By applying spectral techniques, the data becomes easier to cluster and then k-means is classically used. Unfortunately the number of clusters must be manually set and it is very sensitive to initialization. Moreover, k-means tends to split large clusters, to merge small clusters, and to favor convex-shaped clusters. In this ...
متن کاملDoubly Stochastic Normalization for Spectral Clustering
In this paper we focus on the issue of normalization of the affinity matrix in spectral clustering. We show that the difference between N-cuts and Ratio-cuts is in the error measure being used (relative-entropy versus L1 norm) in finding the closest doubly-stochastic matrix to the input affinity matrix. We then develop a scheme for finding the optimal, under Frobenius norm, doubly-stochastic ap...
متن کامل