Abstract
Given a set of data, one central goal is to group them into clusters based on some notion of similarity between the individual objects. One of the most popular and widely-used approaches is k-means despite the computational hardness to find its global minimum. We study and compare the properties of different convex relaxations by relating them to corresponding proximity conditions, an idea originally introduced by Kumar and Kannan. Using conic duality theory, we present an improved proximity condition under which the Peng–Wei relaxation of k-means recovers the underlying clusters exactly. Our proximity condition improves upon Kumar and Kannan and is comparable to that of Awashti and Sheffet, where proximity conditions are established for projective k-means. In addition, we provide a necessary proximity condition for the exactness of the Peng–Wei relaxation. For the special case of equal cluster sizes, we establish a different and completely localized proximity condition under which the Amini–Levina relaxation yields exact clustering, thereby having addressed an open problem by Awasthi and Sheffet in the balanced case. Our framework is not only deterministic and model-free but also comes with a clear geometric meaning which allows for further analysis and generalization. Moreover, it can be conveniently applied to analyzing various data generative models such as the stochastic ball models and Gaussian mixture models. With this method, we improve the current minimum separation bound for the stochastic ball models and achieve the state-of-the-art results of learning Gaussian mixture models.
Similar content being viewed by others
Notes
\(\mathcal{K}\) is pointed if for \(\varvec{Z}\in \mathcal{K}\) and \(-\varvec{Z}\in \mathcal{K}\), \(\varvec{Z}\) must be \(\varvec{0}\), see Chapter 2 in [7].
The dual cone of \(\mathcal{K}\) is defined as \(\{\varvec{W}: \langle \varvec{W}, \varvec{Z}\rangle \ge 0, \forall \varvec{Z}\in \mathcal{K}\}\); in particular, there holds \((\mathcal{K}^*)^* = \mathcal{K}.\)
The primal problem or dual problem is solvable if it is feasible, bounded and the optimal value is attained.
References
Achlioptas, D., McSherry, F.: On spectral learning of mixtures of distributions. In: International Conference on Computational Learning Theory, pp. 458–469. Springer, New York (2005)
Aloise, D., Deshpande, A., Hansen, P., Popat, P.: NP-hardness of Euclidean sum-of-squares clustering. Mach. Learn. 75(2), 245–248 (2009)
Amini, A.A., Levina, E.: On semidefinite relaxations for the block model. Ann. Stat. 46(1), 149–179 (2018)
Arthur, D., Manthey, B., Röglin, H.: Smoothed analysis of the k-means method. J. ACM (JACM) 58(5), 19 (2011)
Awasthi, P., Bandeira, A.S., Charikar, M., Krishnaswamy, R., Villar, S., Ward, R.: Relax, no need to round: integrality of clustering formulations. In: Proceedings of the 2015 Conference on Innovations in Theoretical Computer Science, pp. 191–200. ACM (2015)
Awasthi, P., Sheffet, O.: Improved spectral-norm bounds for clustering. In: APPROX-RANDOM, pp. 37–49. Springer, New York (2012)
Ben-Tal, A., Nemirovski, A.: Lectures on modern convex optimization: analysis, algorithms, and engineering applications. SIAM (2001)
Boyd, S., Vandenberghe, L.: Convex Optimization. Cambridge University Press, Cambridge (2004)
Dasgupta, S.: Learning mixtures of gaussians. In: 40th Annual Symposium on Foundations of Computer Science, pp. 634–644. IEEE (1999)
Du, Q., Faber, V., Gunzburger, M.: Centroidal Voronoi tessellations: applications and algorithms. SIAM Rev. 41(4), 637–676 (1999)
Golub, G.H., Van Loan, C.F.: Matrix Computations, 3rd edn. The Johns Hopkins University Press, Baltimore (1996)
Iguchi, T., Mixon, D.G., Peterson, J., Villar, S.: On the tightness of an SDP relaxation of k-means (2015). arXiv preprint arXiv:1505.04778
Iguchi, T., Mixon, D.G., Peterson, J., Villar, S.: Probably certifiably correct k-means clustering. Math. Progr. 165(2), 605–642 (2017)
Kannan, R., Vempala, S.: Spectral algorithms. Found. Trends Theor. Comput. Sci. 4(3–4), 157–288 (2009)
Kumar, A., Kannan, R.: Clustering with spectral norm and the k-means algorithm. In: 2010 51st Annual IEEE Symposium on Foundations of Computer Science (FOCS), pp. 299–308. IEEE (2010)
Ling, S., Strohmer, T.: Certifying global optimality of graph cuts via semidefinite relaxation: a performance guarantee for spectral clustering (2018). arXiv preprint arXiv:1806.11429
Lloyd, S.: Least squares quantization in PCM. IEEE Trans. Inf. Theory 28(2), 129–137 (1982)
Lu, Y., Zhou, H.H.: Statistical and computational guarantees of Lloyd’s algorithm and its variants (2016). arXiv preprint arXiv:1612.02099
Mahajan, M., Nimbhorkar, P., Varadarajan, K.: The planar k-means problem is NP-hard. In: International Workshop on Algorithms and Computation, pp. 274–285. Springer, New York (2009)
Mixon, D.G., Villar, S., Ward, R.: Clustering subgaussian mixtures by semidefinite programming. Inf. Inference: J. IMA 6(4), 389–415 (2017)
Peng, J., Wei, Y.: Approximating k-means-type clustering via semidefinite programming. SIAM J. Opt. 18(1), 186–205 (2007)
Selim, S.Z., Ismail, M.A.: k-Means-type algorithms: a generalized convergence theorem and characterization of local optimality. IEEE Trans. Pattern Anal. Mach. Intell. 6(1), 81–87 (1984)
Tropp, J.A.: User-friendly tail bounds for sums of random matrices. Found. Comput. Math. 12(4), 389–434 (2012)
Vattani, A.: k-Means requires exponentially many iterations even in the plane. Discrete Comput. Geom. 45(4), 596–616 (2011)
Vempala, S., Wang, G.: A spectral algorithm for learning mixture models. J. Comput. Syst. Sci. 68(4), 841–860 (2004)
Vershynin, R.: Introduction to the non-asymptotic analysis of random matrices. In: Eldar, Y.C., Kutyniok, G. (eds.) Compressed Sensing: Theory and Applications, Chapter 5. Cambridge University Press, Cambridge (2012)
Wright, S.J.: Primal-dual interior-point methods. SIAM (1997)
Yang, L., Sun, D., Toh, K.-C.: SDPNAL+: a majorized semismooth Newton-CG augmented Lagrangian method for semidefinite programming with nonnegative constraints. Math. Progr. 7(3), 331–366 (2015)
Zhao, X.-Y., Sun, D., Toh, K.-C.: A Newton-CG augmented Lagrangian method for semidefinite programming. SIAM J. Opt. 20(4), 1737–1765 (2010)
Author information
Authors and Affiliations
Corresponding author
Additional information
Y. Li, S. Ling, T. Strohmer, and K. Wei acknowledge support from the NSF via Grants DMS 1620455 and DMS 1737943.
Rights and permissions
About this article
Cite this article
Li, X., Li, Y., Ling, S. et al. When do birds of a feather flock together? k-Means, proximity, and conic programming. Math. Program. 179, 295–341 (2020). https://doi.org/10.1007/s10107-018-1333-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10107-018-1333-x