A Proximal Approach for a Class of Matrix Optimization Problems

Abstract

In recent years, there has been a growing interest in mathematical models leading to the minimization, in a symmetric matrix space, of a Bregman divergence coupled with a regularization term. We address problems of this type within a general framework where the regularization term is split in two parts, one being a spectral function while the other is arbitrary. A Douglas-Rachford approach is proposed to address such problems and a list of proximity operators is provided allowing us to consider various choices for the fit-to-data functional and for the regularization term. Numerical experiments show the validity of this approach for solving convex optimization problems encountered in the context of sparse covariance matrix estimation. Based on our theoretical results, an algorithm is also proposed for noisy graphical lasso where a precision matrix has to be estimated in the presence of noise. The nonconvexity of the resulting objective function is dealt with a majorization-minimization approach, i.e. by building a sequence of convex surrogates and solving the inner optimization subproblems via the aforementioned Douglas-Rachford procedure. We establish conditions for the convergence of this iterative scheme and we illustrate its good numerical performance with respect to state-of-the-art approaches.

References in corpus (6)

• Proceedings of the 29th International Conference on Machine Learning (ICML-12)
• Sinc approximation of algebraically decaying functions
• Sparse permutation invariant covariance estimation
• Projected Subgradient Methods for Learning Sparse Gaussians
• Sparse Inverse Covariance Selection via Alternating Linearization Methods
• Gaussian Graphical Models: An Algebraic and Geometric Perspective

Cited by in corpus (6)

• Fixed Point Strategies in Data Science
• Graphical Inference in Linear-Gaussian State-Space Models
• A Laplacian Approach to ${\ell }_1$-Norm Minimization
• Sparse Graphical Linear Dynamical Systems
• Efficient Hierarchical Bayesian Inference for Spatio-temporal Regression Models in Neuroimaging
• Implicit Differentiation for Hyperparameter Tuning the Weighted Graphical Lasso