Corpus ID: 3259

Spectral Clustering of Graphs with General Degrees in the Extended Planted Partition Model

  title={Spectral Clustering of Graphs with General Degrees in the Extended Planted Partition Model},
  author={Kamalika Chaudhuri and Fan Chung Graham and Alexander Tsiatas},
In this paper, we examine a spectral clustering algorithm for similarity graphs drawn from a simple random graph model, where nodes are allowed to have varying degrees, and we provide theoretical bounds on its performance. The random graph model we study is the Extended Planted Partition (EPP) model, a variant of the classical planted partition model. The standard approach to spectral clustering of graphs is to compute the bottom k singular vectors or eigenvectors of a suitable graph Laplacian… Expand
Improved Graph Clustering
This paper presents a new algorithm-a convexified version of maximum likelihood-for graph clustering, and shows that, in the classic stochastic block model setting, it outperforms existing methods by polynomial factors when the cluster size is allowed to have general scalings. Expand
Diffusion and clustering on large graphs
This dissertation studies two important algorithmic problems on networks: graph diffusion and clustering. These problems are closely related: bottlenecks that diffusive processes find difficult toExpand
Title Concentration and regularization of random graphs Permalink
This paper studies how close random graphs are typically to their expectations. We interpret this question through the concentration of the adjacency and Laplacian matrices in the spectral norm. WeExpand
Understanding Regularized Spectral Clustering via Graph Conductance
The results show that unbalanced partitions from spectral clustering can be understood as overfitting to noise in the periphery of a sparse and stochastic graph and demonstrate how regularization can improve the computational speed of spectral clusters. Expand
Challenges in random graph models with degree heterogeneity: existence, enumeration and asymptotics of the spectral radius
In order to understand how the network structure impacts the underlying dynamics, we seek an assortment of methods for efficiently constructing graphs of interest that resemble their empiricallyExpand
Concentration and regularization of random graphs
This paper studies how close random graphs are typically to their expectations through the concentration of the adjacency and Laplacian matrices in the spectral norm, and builds a new decomposition of random graphs based on Grothendieck-Pietsch factorization. Expand
Sparse random graphs: regularization and concentration of the Laplacian
It is proved that this regularization indeed forces Laplacian to concentrate even in sparse graphs, establishing the validity of one of the simplest and fastest approaches to community detection -- regularized spectral clustering, under the stochastic block model. Expand
Asymptotics of the spectral radius for directed Chung-Lu random graphs with community structure
The spectral radius of the adjacency matrix can impact both algorithmic efficiency as well as the stability of solutions to an underlying dynamical process and this work provides novel concentration results for the spectral radius for the directed Chung-Lu random graph model. Expand
Searching for a Single Community in a Graph
This article develops a variant of the method of moments that identifies nodes in the target more reliably, and with lower computation, than generic community detection methods that do not use side information and partition the entire graph. Expand
Regularized spectral methods for clustering signed networks
This work builds on the results in [CDGT 2019] on two fronts for the normalized versions of SPONGE and the Signed Laplacian, and introduces regularized versions of both methods to handle sparse graphs in the regime where the graph is moderately dense. Expand


Finding Planted Partitions in Random Graphs with General Degree Distributions
There is a polynomial time algorithm for recovering (a large part of) the planted partition in this model even in the sparse case, where the average degree is constant, and this model is characterized by a prescribed expected degree sequence. Expand
Spectral analysis of random graphs with skewed degree distributions
This work extends spectral methods to random graphs with skewed degree distributions through a degree based normalization closely connected to the normalized Laplacian, and proves that after applying the transformation, spectral analysis succeeds in recovering the latent structure with high probability. Expand
Concentration of the adjacency matrix and of the Laplacian in random graphs with independent edges
Consider any random graph model where potential edges appear independently, with possibly different probabilities, and assume that the minimum expected degree is !(lnn). We prove that the adjacencyExpand
Algorithms for graph partitioning on the planted partition model
The NP-hard graph bisection problem is to partition the nodes of an undirected graph into two equal-sized groups so as to minimize the number of edges that cross the partition. The more general graphExpand
Spectral graph theory
  • U. Feige
  • Zeta and 𝐿-functions in Number Theory and Combinatorics
  • 2019
With every graph (or digraph) one can associate several different matrices. We have already seen the vertex-edge incidence matrix, the Laplacian and the adjacency matrix of a graph. Here we shallExpand
Spectral partitioning of random graphs
  • Frank McSherry
  • Computer Science
  • Proceedings 2001 IEEE International Conference on Cluster Computing
  • 2001
This paper shows that a simple spectral algorithm can solve all three problems above in the average case, as well as a more general problem of partitioning graphs based on edge density. Expand
Finding Planted Partitions in Nearly Linear Time using Arrested Spectral Clustering
Lower bounds are provided that imply that the "large enough" constraint cannot be improved much, even using a computationally unbounded algorithm. Expand
Spectral clustering and the high-dimensional stochastic blockmodel
Networks or graphs can easily represent a diverse set of data sources that are characterized by interacting units or actors. Social ne tworks, representing people who communicate with each other, areExpand
On the Spectra of General Random Graphs
A Chernoff inequality for matrices is used to show that the eigenvalues of the adjacency matrix and the normalized Laplacian of such a random graph can be approximated by those of the weighted expectation graph, with error bounds dependent upon the minimum and maximum expected degrees. Expand
Complex Graphs and Networks
Graph theory in the information age Old and new concentration inequalities A generative model--the preferential attachment scheme Duplication models for biological networks Random graphs with givenExpand