Affiliation:
1. Mississippi State University, Starkville, MS 39762, USA
Abstract
We investigate a class of challenging general semidefinite programming problems with extra nonconvex constraints such as matrix rank constraints. This problem has extensive applications, including combinatorial graph problems, such as MAX-CUT and community detection, reformulated as quadratic objectives over nonconvex constraints. A customized approach based on the alternating direction method of multipliers (ADMM) is proposed to solve the general large-scale nonconvex semidefinite programming efficiently. We propose two reformulations: one using vector variables and constraints, and the other further reformulating the Burer–Monteiro form. Both formulations admit simple subproblems and can lead to significant improvement in scalability. Despite the nonconvex constraint, we prove that the ADMM iterates converge to a stationary point in both formulations, under mild assumptions. Additionally, recent work suggests that in this matrix form, when the matrix factors are wide enough, the local optimum with high probability is also the global optimum. To demonstrate the scalability of our algorithm, we include results for MAX-CUT, community detection, and image segmentation.
Subject
General Mathematics,Engineering (miscellaneous),Computer Science (miscellaneous)
Reference75 articles.
1. Bandeira, A.S., Boumal, N., and Voroninski, V. (2016, January 23–26). On the low-rank approach for semidefinite programs arising in synchronization and community detection. Proceedings of the Conference on Learning Theory, New York, NY, USA.
2. Community detection in networks: A user guide;Fortunato;Phys. Rep.,2016
3. Phase transitions in semidefinite relaxations;Javanmard;Proc. Natl. Acad. Sci. USA,2016
4. Gillis, N. (2011). Nonnegative Matrix Factorization: Complexity, Algorithms and Applications. [Doctoral Dissertation, Université Catholique de Louvain (CORE)].
5. Ding, C., He, X., and Simon, H.D. (2005, January 11–15). On the equivalence of nonnegative matrix factorization and spectral clustering. Proceedings of the 2005 SIAM International Conference on Data Mining (SIAM), New Orleans, LA, USA.