Spectral Graph Matching and Regularized Quadratic Relaxations I: The Gaussian Model

Graph matching aims at finding the vertex correspondence between two unlabeled graphs that maximizes the total edge weight correlation. This amounts to solving a computationally intractable quadratic assignment problem. In this paper we propose a new spectral method, GRAph Matching by Pairwise eigen-Alignments (GRAMPA). Departing from prior spectral approaches that only compare top eigenvectors, or eigenvectors of the same order, GRAMPA first constructs a similarity matrix as a weighted sum of outer products between all pairs of eigenvectors of the two graphs, with weights given by a Cauchy kernel applied to the separation of the corresponding eigenvalues, then outputs a matching by a simple rounding procedure. The similarity matrix can also be interpreted as the solution to a regularized quadratic programming relaxation of the quadratic assignment problem. For the Gaussian Wigner model in which two complete graphs on $n$ vertices have Gaussian edge weights with correlation coefficient $1-\sigma^2$, we show that GRAMPA exactly recovers the correct vertex correspondence with high probability when $\sigma = O(\frac{1}{\log n})$. This matches the state of the art of polynomial-time algorithms, and significantly improves over existing spectral methods which require $\sigma$ to be polynomially small in $n$. The superiority of GRAMPA is also demonstrated on a variety of synthetic and real datasets, in terms of both statistical accuracy and computational efficiency. Universality results, including similar guarantees for dense and sparse Erd\H{o}s-R\'{e}nyi graphs, are deferred to the companion paper.

[1]  Alexandre d'Aspremont,et al.  Convex Relaxations for Permutation Problems , 2013, SIAM J. Matrix Anal. Appl..

[2]  M. Zaslavskiy,et al.  A Path Following Algorithm for the Graph Matching Problem , 2008, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  G. Finke Quadratic Assignment Problems , 1987 .

[4]  F. Gotze,et al.  On the Rate of Convergence to the Marchenko--Pastur Distribution , 2011, 1110.1284.

[5]  Matthias Grossglauser,et al.  Growing a Graph Matching from a Handful of Seeds , 2015, Proc. VLDB Endow..

[6]  Matthias Grossglauser,et al.  On the Structure and Efficient Computation of IsoRank Node Similarities , 2016, ArXiv.

[7]  Alexander Tikhomirov,et al.  On the rate of convergence to the semi-circular law , 2011, 1109.0611.

[8]  Roman Vershynin,et al.  High-Dimensional Probability , 2018 .

[9]  Elchanan Mossel,et al.  Seeded graph matching via large neighborhood statistics , 2018, SODA.

[10]  Daniel Cullina,et al.  Improved Achievability and Converse Bounds for Erdos-Renyi Graph Matching , 2016, SIGMETRICS.

[11]  Maxim Sviridenko,et al.  Concentration and moment inequalities for polynomials of independent random variables , 2012, SODA.

[12]  L. Benigni Eigenvectors distribution and quantum unique ergodicity for deformed Wigner matrices , 2017, Annales de l'Institut Henri Poincaré, Probabilités et Statistiques.

[13]  Salih O. Duffuaa,et al.  A Linear Programming Approach for the Weighted Graph Matching Problem , 1993, IEEE Trans. Pattern Anal. Mach. Intell..

[14]  Muriel Médard,et al.  Spectral Alignment of Networks , 2015, ArXiv.

[15]  Jiaming Xu,et al.  Efficient random graph matching via degree profiles , 2018, Probability Theory and Related Fields.

[16]  Guillermo Sapiro,et al.  Graph Matching: Relax at Your Own Risk , 2014, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[17]  A. Guionnet,et al.  An Introduction to Random Matrices , 2009 .

[18]  Jiaming Xu,et al.  Spectral Graph Matching and Regularized Quadratic Relaxations II: Erdős-Rényi Graphs and Universality , 2019, Foundations of Computational Mathematics.

[19]  M. Rudelson,et al.  Hanson-Wright inequality and sub-gaussian concentration , 2013 .

[20]  Silvio Lattanzi,et al.  An efficient reconciliation algorithm for social networks , 2013, Proc. VLDB Endow..

[21]  Carey E. Priebe,et al.  Fast Approximate Quadratic Programming for Graph Matching , 2015, PloS one.

[22]  Daniel Cullina,et al.  Exact alignment recovery for correlated Erdos Renyi graphs , 2017, ArXiv.

[23]  Yaron Lipman,et al.  DS++ , 2017, ACM Trans. Graph..

[24]  Bonnie Berger,et al.  Global alignment of multiple protein interaction networks with application to functional orthology detection , 2008, Proceedings of the National Academy of Sciences.

[25]  Christoph Schnörr,et al.  Probabilistic Subgraph Matching Based on Convex Relaxation , 2005, EMMCVPR.

[26]  Daniel Cullina,et al.  Database Alignment with Gaussian Features , 2019, AISTATS.

[27]  P. Erdos,et al.  On the evolution of random graphs , 1984 .

[28]  Carey E. Priebe,et al.  Seeded graph matching for correlated Erdös-Rényi graphs , 2014, J. Mach. Learn. Res..

[29]  Matthias Grossglauser,et al.  On the privacy of anonymized networks , 2011, KDD.

[30]  Lorenzo Livi,et al.  The graph matching problem , 2012, Pattern Analysis and Applications.

[31]  Jure Leskovec,et al.  {SNAP Datasets}: {Stanford} Large Network Dataset Collection , 2014 .

[32]  Gopal Pandurangan,et al.  Improved Random Graph Isomorphism Tomek Czajka , 2006 .

[33]  Mario Vento,et al.  Thirty Years Of Graph Matching In Pattern Recognition , 2004, Int. J. Pattern Recognit. Artif. Intell..

[34]  B. Bollobás The evolution of random graphs , 1984 .

[35]  Y. Aflalo,et al.  On convex relaxation of graph isomorphism , 2015, Proceedings of the National Academy of Sciences.

[36]  H. Yau,et al.  The Eigenvector Moment Flow and Local Quantum Unique Ergodicity , 2013, 1312.1301.

[37]  B. Bollobás Distinguishing Vertices of Random Graphs , 1982 .

[38]  Maxim Sviridenko,et al.  Maximum Quadratic Assignment Problem: Reduction from Maximum Label Cover and LP-based Approximation Algorithm , 2010, TALG.

[39]  Albert,et al.  Emergence of scaling in random networks , 1999, Science.

[40]  Vitaly Shmatikov,et al.  Robust De-anonymization of Large Sparse Datasets , 2008, 2008 IEEE Symposium on Security and Privacy (sp 2008).

[41]  R. Srikant,et al.  Correcting the Output of Approximate Graph Matching Algorithms , 2018, IEEE INFOCOM 2018 - IEEE Conference on Computer Communications.

[42]  Christos Faloutsos,et al.  Graphs over time: densification laws, shrinking diameters and possible explanations , 2005, KDD '05.

[43]  Tselil Schramm,et al.  (Nearly) Efficient Algorithms for the Graph Matching Problem on Correlated Random Graphs , 2018, NeurIPS.

[44]  Yongtang Shi,et al.  Fifty years of graph matching, network alignment and network comparison , 2016, Inf. Sci..

[45]  Prateek Mittal,et al.  Partial Recovery of Erd\H{o}s-R\'{e}nyi Graph Alignment via $k$-Core Alignment , 2018 .

[46]  Panos M. Pardalos,et al.  Quadratic Assignment Problem , 1997, Encyclopedia of Optimization.

[47]  Kathryn B. Laskey,et al.  Stochastic blockmodels: First steps , 1983 .

[48]  Matthias Grossglauser,et al.  PROPER: global protein interaction network alignment through percolation matching , 2016, BMC Bioinformatics.

[49]  Elza Erkip,et al.  Seeded graph matching: Efficient algorithms and theoretical guarantees , 2017, 2017 51st Asilomar Conference on Signals, Systems, and Computers.

[50]  Panos M. Pardalos,et al.  The Quadratic Assignment Problem: A Survey and Recent Developments , 1993, Quadratic Assignment and Related Problems.

[51]  Lei Xu,et al.  A PCA approach for fast retrieval of structural patterns in attributed graphs , 2001, IEEE Trans. Syst. Man Cybern. Part B.

[52]  Osman Emre Dai,et al.  Analysis of a Canonical Labeling Algorithm for the Alignment of Correlated Erdős-Rényi Graphs , 2018, Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems.

[53]  Muriel Medard,et al.  Spectral Alignment of Graphs , 2016, IEEE Transactions on Network Science and Engineering.

[54]  H. Kuhn The Hungarian method for the assignment problem , 1955 .

[55]  Shinji Umeyama,et al.  An Eigendecomposition Approach to Weighted Graph Matching Problems , 1988, IEEE Trans. Pattern Anal. Mach. Intell..

[56]  Matthias Grossglauser,et al.  On the performance of percolation graph matching , 2013, COSN '13.

[57]  Daniel Cullina,et al.  On the Performance of a Canonical Labeling for Matching Correlated Erdős-Rényi Graphs , 2018, ArXiv.

[58]  Vitaly Shmatikov,et al.  De-anonymizing Social Networks , 2009, 2009 30th IEEE Symposium on Security and Privacy.

[59]  Tosio Kato Continuity of the Map S→|S| for Linear Operators , 1973 .

[60]  David M. Mount,et al.  Isomorphism of graphs with bounded eigenvalue multiplicity , 1982, STOC '82.