Matrix completion via a low rank factorization model and an Augmented Lagrangean Succesive Overrelaxation Algorithm
Journal Title: Bulletin of Computational Applied Mathematics (Bull CompAMa) - Year 2014, Vol 2, Issue 2
Abstract
The matrix completion problem (MC) has been approximated by using the nuclear norm relaxation. Some algorithms based on this strategy require the computationally expensive singular value decomposition (SVD) at each iteration. One way to avoid SVD calculations is to use alternating methods, which pursue the completion through matrix factorization with a low rank condition. In this work an augmented Lagrangean-type alternating algorithm is proposed. The new algorithm uses duality information to define the iterations, in contrast to the solely primal LMaFit algorithm, which employs a Successive Over Relaxation scheme. The convergence result is studied. Some numerical experiments are given to compare numerical performance of both proposals.
Authors and Affiliations
Hugo Lara, Harry Oviedo, Jinjun Yuan
Keywords
Related Articles
- EP ID EP245119
- DOI -
- Views 81
- Downloads 0
New results on the stability, integrability and boundedness in Volterra integro-differential equations
The authors of this article deal with a first order non-linear Volterra integro-differential equation (NVIDE). To this end, the conditions are obtained which are sufficient for stability (S), boundedness (B), and for eve...
Redox reactions as experimental examples of ternary weak algebraic hyperstructures
A ternary hyperoperation on a set H is a 3-ary hyperoperation, which associates a subset of H with any three elements of H. In this paper, we give examples of ternary hyperoperations associated with redox reactions. We...
(Free) Software for general partial differential equation problems in non-rectangular 2D and 3D regions
PDE2D is a general-purpose partial differential equation solver which solves very general systems of nonlinear, steady-state, time-dependent and eigenvalue PDEs in 1D intervals, general 2D regions (see Figure 1), and a w...
The exponential distribution as the sum of discontinuous distributions
We show that for any natural number $n$, an exponential distribution can be written as the sum of $n$ discontinuous variables and another exponential distribution, all of them independent.
Constrained optimization with integer and continuous variables using inexact restoration and projected gradients
Inexact restoration (IR) is a well established technique for continuous minimization problems with constraints that can be applied to constrained optimization problems with specific structures. When some variables are re...