We consider a multiple-block separable convex programming problem, where the objective function is the sum of m individual convex functions without overlapping variables, and the constraints are linear, aside from sid...
详细信息
We consider a multiple-block separable convex programming problem, where the objective function is the sum of m individual convex functions without overlapping variables, and the constraints are linear, aside from side constraints. Based on the combination of the classical Gauss-Seidel and the Jacobian decompositions of the augmented Lagrangian function, we propose a partially parallel splitting method, which differs from existing augmented Lagrangian based splittingmethods in the sense that such an approach simplifies the iterative scheme significantly by removing the potentially expensive correction step. Furthermore, a relaxation step, whose computational cost is negligible, can be incorporated into the proposed method to improve its practical performance. Theoretically, we establish global convergence of the new method in the framework of proximal point algorithm and worst-case nonasymptotic O(1/t) convergence rate results in both ergodic and nonergodic senses, where t counts the iteration. The efficiency of the proposed method is further demonstrated through numerical results on robust PCA, i.e., factorizing from incomplete information of an unknown matrix into its low-rank and sparse components, with both synthetic and real data of extracting the background from a corrupted surveillance video.
A popular optimization model arising from image processing is the separable optimization problem whose objective function is the sum of three independent functions, and the variables are coupled by a linear equality. ...
详细信息
A popular optimization model arising from image processing is the separable optimization problem whose objective function is the sum of three independent functions, and the variables are coupled by a linear equality. In this paper, we propose to solve such problem by a new partially parallel splitting method, whose step sizes for the primal and the dual variables in correction step are not necessarily identical. We establish the global convergence, and study the convergence rate on this varying ADMM-based prediction-correction method named as VAPCM. We derive the worst-case O(1/t) convergence rate in both the ergodic and non-ergodic senses. We also show that the convergence rate can be improved to o(1/t). Moreover, under the error bound assumptions, we establish the global linear convergence of VAPCM. We apply the new method to solve problems in robust principal component analysis and image decomposition. Numerical results indicate that the new method is efficient.
暂无评论