The Annals of Statistics

Graph connection Laplacian methods can be made robust to noise

Noureddine El Karoui and Hau-Tieng Wu

Full-text: Open access


Recently, several data analytic techniques based on graph connection Laplacian (GCL) ideas have appeared in the literature. At this point, the properties of these methods are starting to be understood in the setting where the data is observed without noise. We study the impact of additive noise on these methods and show that they are remarkably robust. As a by-product of our analysis, we propose modifications of the standard algorithms that increase their robustness to noise. We illustrate our results in numerical simulations.

Article information

Ann. Statist., Volume 44, Number 1 (2016), 346-372.

Received: May 2014
Revised: September 2014
First available in Project Euclid: 5 January 2016

Permanent link to this document

Digital Object Identifier

Mathematical Reviews number (MathSciNet)

Zentralblatt MATH identifier

Primary: 60F99: None of the above, but in this section
Secondary: 53A99: None of the above, but in this section

Concentration of measure random matrices graph connection Laplacian vector diffusion maps spectral geometry kernel methods


El Karoui, Noureddine; Wu, Hau-Tieng. Graph connection Laplacian methods can be made robust to noise. Ann. Statist. 44 (2016), no. 1, 346--372. doi:10.1214/14-AOS1275.

Export citation


  • [1] Alexeev, B., Bandeira, A. S., Fickus, M. and Mixon, D. G. (2014). Phase retrieval with polarization. SIAM J. Imaging Sci. 7 35–66.
  • [2] Bandeira, A. S., Singer, A. and Spielman, D. A. (2013). A Cheeger inequality for the graph connection Laplacian. SIAM J. Matrix Anal. Appl. 34 1611–1630.
  • [3] Batard, T. and Sochen, N. (2012). Polyakov action on ($\rho,G$)-equivariant functions application to color image regularization. In Scale Space and Variational Methods in Computer Vision 483–494. Springer, Berlin.
  • [4] Batard, T. and Sochen, N. (2014). A class of generalized Laplacians on vector bundles devoted to multi-channel image processing. J. Math. Imaging Vision 48 517–543.
  • [5] Bhatia, R. (1997). Matrix Analysis. Graduate Texts in Mathematics 169. Springer, New York.
  • [6] Boyer, D. M., Lipman, Y., Clair, E. S., Puente, J., Patel, B. A., Funkhouser, T., Jernvall, J. and Daubechies, I. (2011). Algorithms to automatically quantify the geometric similarity of anatomical surfaces. Proc. Natl. Acad. Sci. USA 108 18221–18226.
  • [7] Chen, P., Lin, C.-L. and Chern, I.-L. (2013). A perfect match condition for point-set matching problems using the optimal mass transport approach. SIAM J. Imaging Sci. 6 730–764.
  • [8] Chung, F. and Kempton, M. (2013). A local clustering algorithm for connection graphs. In Algorithms and Models for the Web Graph (A. Bonato, M. Mitzenmacher and P. Pralat, eds.). Lecture Notes in Computer Science 8305 26–43. Springer, Berlin.
  • [9] Chung, F., Zhao, W. and Kempton, M. (2012). Ranking and sparsifying a connection graph. In Algorithms and Models for the Web Graph (A. Bonato and J. Janssen, eds.). Lecture Notes in Computer Science 7323 66–77. Springer, Berlin.
  • [10] Collins, A., Zomorodian, A., Carlsson, G. and Guibas, L. J. (2004). A barcode shape descriptor for curve point cloud data. Computers & Graphics 28 881–894.
  • [11] Cucuringu, M., Lipman, Y. and Singer, A. (2012). Sensor network localization by eigenvector synchronization over the Euclidean group. ACM Trans. Sens. Netw. 8 19:1–19:42.
  • [12] Cucuringu, M., Singer, A. and Cowburn, D. (2012). Eigenvector synchronization, graph rigidity and the molecule problem. Inf. Inference 1 21–67.
  • [13] Diaconis, P. and Freedman, D. (1984). Asymptotics of graphical projection pursuit. Ann. Statist. 12 793–815.
  • [14] El Karoui, N. (2008). Operator norm consistent estimation of large-dimensional sparse covariance matrices. Ann. Statist. 36 2717–2756.
  • [15] El Karoui, N. (2009). Concentration of measure and spectra of random matrices: Applications to correlation matrices, elliptical distributions and beyond. Ann. Appl. Probab. 19 2362–2405.
  • [16] El Karoui, N. (2010). On information plus noise kernel random matrices. Ann. Statist. 38 3191–3216.
  • [17] El Karoui, N., Bean, D., Bickel, P. J., Lim, C. and Yu, B. (2013). On robust regression with high-dimensional predictors. Proc. Natl. Acad. Sci. USA 110 14557–14562.
  • [18] El Karoui, N. and Wu, H.-T. (2013). Vector diffusion maps and random matrices with random blocks. Available at arXiv:1310.0188.
  • [19] El Karoui, N. and Wu, H. (2015). Supplement to “Graph connection Laplacian methods can be made robust to noise.” DOI:10.1214/14-AOS1275SUPP.
  • [20] Epstein, C. L. (2008). Introduction to the Mathematics of Medical Imaging, 2nd ed. SIAM, Philadelphia, PA.
  • [21] Fagin, R., Kumar, R. and Sivakumar, D. (2003). Comparing top $k$ lists. SIAM J. Discrete Math. 17 134–160 (electronic).
  • [22] Hadani, R. and Singer, A. (2011). Representation theoretic patterns in three dimensional cryo-electron microscopy I: The intrinsic reconstitution algorithm. Ann. of Math. (2) 174 1219–1241.
  • [23] Hall, P., Marron, J. S. and Neeman, A. (2005). Geometric representation of high dimension, low sample size data. J. R. Stat. Soc. Ser. B Stat. Methodol. 67 427–444.
  • [24] Huang, Q.-X., Su, H. and Guibas, L. (2013). Fine-grained semi-supervised labeling of large shape collections. ACM Transactions on Graphics (TOG) 32 190.
  • [25] Ledoux, M. (2001). The Concentration of Measure Phenomenon. Mathematical Surveys and Monographs 89. Amer. Math. Soc., Providence, RI.
  • [26] Lipman, Y., Al-Aifari, R. and Daubechies, I. (2013). Continuous Procrustes distance between two surfaces. Comm. Pure Appl. Math. 66 934–964.
  • [27] Marchesini, S., Tu, Y.-C. and Wu, H.-T. (2015). Alternating projection, ptychographic imaging and phase synchronization. Appl. Comput. Harmon. Anal. DOI:10.1016/j.acha.2015.06.005. Available at arXiv:1402.0550.
  • [28] Mémoli, F. and Sapiro, G. (2005). A theoretical and computational framework for isometry invariant recognition of point cloud data. Found. Comput. Math. 5 313–347.
  • [29] Niyogi, P., Smale, S. and Weinberger, S. (2009). Finding the homology of submanifolds with high confidence from random samples. In Twentieth Anniversary Volume 1–23. Springer, New York.
  • [30] Singer, A. (2011). Angular synchronization by eigenvectors and semidefinite programming. Appl. Comput. Harmon. Anal. 30 20–36.
  • [31] Singer, A. and Wu, H.-T. (2012). Vector diffusion maps and the connection Laplacian. Comm. Pure Appl. Math. 65 1067–1144.
  • [32] Singer, A. and Wu, H.-T. (2013). Two-dimensional tomography from noisy projections taken at unknown random directions. SIAM J. Imaging Sci. 6 136–175.
  • [33] Singer, A. and Wu, H.-T. (2013). Spectral convergence of the connection laplacian from random samples. Submitted.
  • [34] Singer, A., Zhao, Z., Shkolnisky, Y. and Hadani, R. (2011). Viewing angle classification of cryo-electron microscopy images using eigenvectors. SIAM J. Imaging Sci. 4 723–759.
  • [35] Stewart, G. W. and Sun, J. G. (1990). Matrix Perturbation Theory. Academic Press, Boston, MA.
  • [36] Sun, J., Ovsjanikov, M. and Guibas, L. (2009). A concise and provably informative multi-scale signature based on heat diffusion. In Proceedings of the Symposium on Geometry Processing, SGP ’09 1383–1392. Eurographics Association.
  • [37] Talmon, R., Cohen, I., Gannot, S. and Coifman, R. (2013). Diffusion maps for signal processing: A deeper look at manifold-learning techniques based on kernels and graphs. IEEE Signal Process. Mag. 30 75–86.
  • [38] van der Vaart, A. W. (1998). Asymptotic Statistics. Cambridge Series in Statistical and Probabilistic Mathematics 3. Cambridge Univ. Press, Cambridge.
  • [39] Wang, F., Huang, Q. and Guibas, L. J. (2013). Image co-segmentation via consistent functional maps. In The IEEE International Conference on Computer Vision (ICCV).
  • [40] Zhao, Z. and Singer, A. (2013). Rotationally invariant image representation for viewing direction classification in cryo-EM. Available at arXiv:1309.7643.

Supplemental materials