James R. Lee
Shayan Oveis Gharan
Abstract
A basic fact in spectral graph theory is that the number of connected components in an undirected graph is equal to the multiplicity of the eigenvalue zero in the Laplacian matrix of the graph. In particular, the graph is disconnected if and only if there are at least two eigenvalues equal to zero. Cheeger's inequality and its variants provide an approximate version of the latter fact; they state that a graph has a sparse cut if and only if there are at least two eigenvalues that are close to zero.
It has been conjectured that an analogous characterization holds for higher multiplicities: There are k eigenvalues close to zero if and only if the vertex set can be partitioned into k subsets, each defining a sparse cut. We resolve this conjecture positively. Our result provides a theoretical justification for clustering algorithms that use the bottom k eigenvectors to embed the vertices into Rk, and then apply geometric considerations to the embedding.
We also show that these techniques yield a nearly optimal quantitative connection between the expansion of sets of size ≈ n/k and λk, the kth smallest eigenvalue of the normalized Laplacian, where n is the number of vertices. In particular, we show that in every graph there are at least k/2 disjoint sets (one of which will have size at most 2n/k), each having expansion at most O(√λk log k). Louis, Raghavendra, Tetali, and Vempala have independently proved a slightly weaker version of this last result. The √log k bound is tight, up to constant factors, for the “noisy hypercube” graphs.
- N. Alon. 1986. Eigenvalues and expanders. Combinatorica 62 (1986), 83--96. Google Scholar
Digital Library
- Noga Alon and Nabil Kahale. 1997. A spectral technique for coloring random 3-colorable graphs. SIAM J. Comput. 26 (1997), 1733--1748. Google ScholarDigital Library
- N. Alon and V. Milman. 1985. Isoperimetric inequalities for graphs, and superconcentrators. J. Combinat. Theory, Ser. B 38, 1 (1985), 73--88.Google ScholarCross Ref
- Sanjeev Arora, Boaz Barak, and David Steurer. 2010. Subexponential algorithms for unique games and related problems. In Proceedings of FOCS. IEEE Computer Society, 563--572. Google ScholarDigital Library
- Bengt Aspvall and John R. Gilbert. 1984. Graph coloring using eigenvalue decomposition. SIAM J. Algeb. Disc. Meth. 5, 4 (1984), 526--538.Google ScholarCross Ref
- Boaz Barak, Fernando G. S. L. Brandão, Aram W. Harrow, Jonathan A. Kelner, David Steurer, and Yuan Zhou. 2012. Hypercontractivity, sum-of-squares proofs, and their applications. In Proceedings of STOC. ACM, 307--326. Google ScholarDigital Library
- William Beckner. 1975. Inequalities in Fourier analysis. Ann. of Math. (2) 102, 1 (1975), 159--182.Google Scholar
- Punyashloka Biswal, James R. Lee, and Satish Rao. 2008. Eigenvalue bounds, spectral partitioning, and metrical deformations via flows. J. ACM 57, 3, (2008), 751--760. Google ScholarDigital Library
- Aline Bonami. 1970. Étude des coefficients de Fourier des fonctions de Lp(G). Ann. Inst. Fourier (Grenoble) 20, fasc. 2 (1970), 335--402.Google ScholarCross Ref
- Sergey Brin and Lawrence Page. 1998. The anatomy of a large-scale hypertextual web search engine. Comput. Netw. 30, 1--7, (1998), 107--117. Google ScholarDigital Library
- M. Charikar, C. Chekuri, A. Goel, and S. Guha. 1998a. Rounding via trees: Deterministic approximation algorithms for group Steiner trees and k-median. In Proceedings of STOC. ACM, 114--123. Google ScholarDigital Library
- Moses Charikar, Chandra Chekuri, Ashish Goel, Sudipto Guha, and Serge A. Plotkin. 1998b. Approximating a finite metric by a small number of tree metrics. In Proceedings of FOCS. IEEE Computer Society, 379--388. Google ScholarDigital Library
- F. R. K. Chung. 1996. Laplacians of graphs and Cheeger's inequalities. In Combinatorics, Paul Erdős Is Eighty, Vol. 2 (Keszthely, 1993). Bolyai Soc. Math. Stud., Vol. 2. János Bolyai Math. Soc., Budapest, 157--172.Google Scholar
- Fan R. K. Chung. 1997. Spectral Graph Theory. CBMS Regional Conference Series in Mathematics, vol. 92. (Published for the Conference Board of the Mathematical Sciences, Washington, DC)Google Scholar
- Amir Daneshgar, Ramin Javadi, and Laurent Miclo. 2012. On nodal domains and higher-order Cheeger inequalities of finite reversible Markov processes. Stochas. Process. Their Applicat. 122, 4 (2012), 1748--1776.Google ScholarCross Ref
- J. Fakcharoenphol and K. Talwar. 2003. An improved decomposition theorem for graphs excluding a fixed minor. In Proceedings of 6th Workshop on Approximation, Randomization, and Combinatorial Optimization Lecture Notes in Computer Science, vol. 2764, Springer, 36--46.Google Scholar
- Anupam Gupta, Robert Krauthgamer, and James R. Lee. 2003. Bounded geometries, fractals, and low-distortion embeddings. In Proceedings of FOCS. IEEE Computer Society, 534--543. Google ScholarDigital Library
- J. Kelner, J. R. Lee, G. Price, and S.-H. Teng. 2011. Metric uniformization and spectral bounds for graphs. Geom. Funct. Anal. 21, 5 (2011), 1117--1143.Google ScholarCross Ref
- Jonathan A. Kelner, Lorenzo Orecchia, Aaron Sidford, and Zeyuan Allen Zhu. 2013. A simple, combinatorial algorithm for solving SDD systems in nearly-linear time. In Proceedings of STOC. ACM, 911--920. Google ScholarDigital Library
- Philip N. Klein, Serge A. Plotkin, and Satish Rao. 1993. Excluded minors, network decomposition, and multicommodity flow. In Proceedings of STOC. ACM, 682--690. Google ScholarDigital Library
- Jon M. Kleinberg. 1999. Authoritative sources in a hyperlinked environment. J. ACM 46, 2 (1999), 668--677. Google ScholarDigital Library
- Ioannis Koutis, Gary L. Miller, and Richard Peng. 2011. A nearly-m log n time solver for SDD linear systems. In Proceedings of FOCS. IEEE Computer Society, 590--598. Google ScholarDigital Library
- Michel Ledoux and Michel Talagrand. 2011. Probability in Banach Spaces: Isoperimetry and Processes, Springer-Verlag, Berlin. Reprint of the 1991 edition.Google Scholar
- James R. Lee and Assaf Naor. 2005. Extending Lipschitz functions via random metric partitions. Invent. Math. 160, 1 (2005), 59--95.Google ScholarCross Ref
- J. R. Lee and A. Sidiropoulos. 2010. Genus and the geometry of the cut graph. In Proceedings of SODA. SIAM, 193--201. Google ScholarDigital Library
- Anand Louis, Prasad Raghavendra, Prasad Tetali, and Santosh Vempala. 2011. Algorithmic extensions of Cheeger's inequality to higher eigenvalues and partitions. In Proceedings of APPROX-RANDOM. Springer, 315--326. Google ScholarDigital Library
- Anand Louis, Prasad Raghavendra, Prasad Tetali, and Santosh Vempala. 2012. Many sparse cuts via higher eigenvalues. In Proceedings of STOC. ACM, 1131--1140. Google ScholarDigital Library
- J. Matoušek. 2002. Lectures on Discrete Geometry. Graduate Texts in Mathematics, vol. 212, Springer-Verlag, NY. Google ScholarDigital Library
- Laurent Miclo. 2008. On eigenfunctions of Markov processes on trees. Probab. Theory Relat. Fields 142, 3--4 (2008), 561--594.Google ScholarCross Ref
- Laurent Miclo. 2013. On hyperboundedness and spectrum of Markov operators. Preprint.Google Scholar
- Andrew Ng, Michael Jordan, and Yair Weiss. 2002. On spectral clustering: Analysis and an algorithm. In Proceedings of NIPS. MIT Press, 849--856.Google Scholar
- Ryan O'Donnell and David Witmer. 2012. Improved small-set expansion from higher eigenvalues. arXiv:1204.4688.Google Scholar
- Shayan Oveis Gharan and Luca Trevisan. 2012. Approximating the expansion profile and almost optimal local graph clustering. In Proceedings of FOCS. IEEE Computer Society, 187--196. Google ScholarDigital Library
- Prasad Raghavendra and David Steurer. 2010. Graph expansion and the unique games conjecture. In Proceedings of STOC. ACM, New York, 755--764. Google ScholarDigital Library
- Prasad Raghavendra, David Steurer, and Prasad Tetali. 2010. Approximations for the isoperimetric and spectral profile of graphs and related parameters. In Proceedings of STOC. ACM, 631--640. Google ScholarDigital Library
- Jianbo Shi and Jitendra Malik. 2000. Normalized cuts and image segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 22, 8 (2000), 888--905. Google ScholarDigital Library
- Barry Simon and Raphael Høegh-Krohn. 1972. Hypercontractive semigroups and two dimensional self-coupled Bose fields. J. Function. Anal. 9 (1972), 121--180.Google ScholarCross Ref
- Alistair J. Sinclair and Mark R. Jerrum. 1989. Approximative counting, uniform generation and rapidly mixing Markov chains. Inf. Computat. 82, 1, (1989), 93--133. Google ScholarDigital Library
- Daniel A. Spielman and Shang-Hua Teng. 2004. Nearly-linear time algorithms for graph partitioning, graph sparsification, and solving linear systems. In Proceedings of STOC. ACM, 81--90. Google ScholarDigital Library
- David Steurer. 2010. On the complexity of unique games and graph expansion. Ph.D. dissertation, Princeton University. Google ScholarDigital Library
- David A. Tolliver and Gary L. Miller. 2006. Graph partitioning by spectral rounding: Applications in image segmentation and clustering. In Proceedings of CVPR. IEEE Computer Society, 1053--1060. Google ScholarDigital Library
- Deepak Verma and Marina Meila. 2003. Comparison of spectral clustering methods. Tech. Rep. UW-CSE-03-05-01. Department of Computer Science, University of Washington.Google Scholar
- Nisheeth K. Vishnoi. 2013. Lx = b. Found. Trends Theoret. Comput. Sci. 8, 1--2 (2013), 1--141.Google ScholarCross Ref
Index Terms
- Multiway Spectral Partitioning and Higher-Order Cheeger Inequalities
Recommendations
Almost Optimal Local Graph Clustering Using Evolving Sets
Spectral partitioning is a simple, nearly linear time algorithm to find sparse cuts, and the Cheeger inequalities provide a worst-case guarantee for the quality of the approximation found by the algorithm. A local graph partitioning algorithm finds a ...
Many sparse cuts via higher eigenvalues
STOC '12: Proceedings of the forty-fourth annual ACM symposium on Theory of computingCheeger's fundamental inequality states that any edge-weighted graph has a vertex subset S such that its expansion (a.k.a. conductance) is bounded as follows: [ φ(S) def= (w(S,bar{S}))/(min set(w(S), w(bar(S)))) ≤ √(2 λ2) ] where w is the total edge ...
Multi-way spectral partitioning and higher-order cheeger inequalities
STOC '12: Proceedings of the forty-fourth annual ACM symposium on Theory of computingA basic fact in spectral graph theory is that the number of connected components in an undirected graph is equal to the multiplicity of the eigenvalue zero in the Laplacian matrix of the graph. In particular, the graph is disconnected if and only if ...
Comments