The construction of an algebraically reduced system for the acceleration of preconditioned conjugate gradients

In this paper we show how an algebraically reduced system can be constructed, for which the preconditioned conjugate gradient method converges faster than for the original system. For this method it is necessary that the original matrix is symmetric positive-definite. Our approach is based on an efficient projection on a well-chosen subspace and we show an application in which a cyclically reduced system is one step further reduced by this novel technique.     

Preconditioned conjugate gradient method for generalized least squares problems

A variant of the preconditioned conjugate gradient method to solve generalized least squares problems is presented. If the problem is min (Ax − b)^TW⁻¹(Ax − b) with A ∈ R^m×n and W ∈ R^m×m symmetric and positive definite, the method needs only a preconditioner A₁ ∈ R^n×n, but not the inverse of matrix W or of any of its submatrices. Freund's comparison result for regular least squares problems is extended to generalized least squares problems. An error bound is also given.     

A nonlinearly preconditioned conjugate gradient algorithm for rank‐R canonical tensor approximation

Alternating least squares (ALS) is often considered the workhorse algorithm for computing the rank‐R canonical tensor approximation, but for certain problems, its convergence can be very slow. The nonlinear conjugate gradient (NCG) method was recently proposed as an alternative to ALS, but the results indicated that NCG is usually not faster than ALS. To improve the convergence speed of NCG, we consider a nonlinearly preconditioned NCG (PNCG) algorithm for computing the rank‐R canonical tensor decomposition. Our approach uses ALS as a nonlinear preconditioner in the NCG algorithm. Alternatively, NCG can be viewed as an acceleration process for ALS. We demonstrate numerically that the convergence acceleration mechanism in PNCG often leads to important pay‐offs for difficult tensor decomposition problems, with convergence that is significantly faster and more robust than for the stand‐alone NCG or ALS algorithms. We consider several approaches for incorporating the nonlinear preconditioner into the NCG algorithm that have been described in the literature previously and have met with success in certain application areas. However, it appears that the nonlinearly PNCG approach has received relatively little attention in the broader community and remains underexplored both theoretically and experimentally. Thus, this paper serves several additional functions, by providing in one place a concise overview of several PNCG variants and their properties that have only been described in a few places scattered throughout the literature, by systematically comparing the performance of these PNCG variants for the tensor decomposition problem, and by drawing further attention to the usefulness of nonlinearly PNCG as a general tool. In addition, we briefly discuss the convergence of the PNCG algorithm. In particular, we obtain a new convergence result for one of the PNCG variants under suitable conditions, building on known convergence results for non‐preconditioned NCG. Copyright © 2014 John Wiley & Sons, Ltd.     

Indefinitely preconditioned conjugate gradient method for large sparse equality and inequality constrained quadratic problems

This paper is concerned with the numerical solution of a symmetric indefinite system which is a generalization of the Karush–Kuhn–Tucker system. Following the recent approach of Luk?an and Vl?ek, we propose to solve this system by a preconditioned conjugate gradient (PCG) algorithm and we devise two indefinite preconditioners with good theoretical properties. In particular, for one of these preconditioners, the finite termination property of the PCG method is stated. The PCG method combined with a parallel version of these preconditioners is used as inner solver within an inexact Interior‐Point (IP) method for the solution of large and sparse quadratic programs. The numerical results obtained by a parallel code implementing the IP method on distributed memory multiprocessor systems enable us to confirm the effectiveness of the proposed approach for problems with special structure in the constraint matrix and in the objective function. Copyright © 2002 John Wiley & Sons, Ltd.     

A conjugate gradient algorithm for sparse linear inequalities

This paper presents a conjugate gradient method for solving systems of linear inequalities. The method is of dual optimization type and consists of two phases which can be implemented in a common framework. Phase 1 either finds the minimum-norm solution of the system or detects the inconsistency of the system. In the latter event, the method proceeds to Phase 2 in which an approximate least-squares solution to the system is obtained. The method is particularly suitable to large scale problems because it preserves the sparsity structure of the problem. Its efficiency is shown by computational comparisons with an SOR type method.     

A scaled nonlinear conjugate gradient algorithm for unconstrained optimization

The best spectral conjugate gradient algorithm by (Birgin, E. and Martínez, J.M., 2001, A spectral conjugate gradient method for unconstrained optimization. Applied Mathematics and Optimization, 43, 117–128). which is mainly a scaled variant of (Perry, J.M., 1977, A class of Conjugate gradient algorithms with a two step varaiable metric memory, Discussion Paper 269, Center for Mathematical Studies in Economics and Management Science, Northwestern University), is modified in such a way as to overcome the lack of positive definiteness of the matrix defining the search direction. This modification is based on the quasi-Newton BFGS updating formula. The computational scheme is embedded into the restart philosophy of Beale–Powell. The parameter scaling the gradient is selected as spectral gradient or in an anticipative way by means of a formula using the function values in two successive points. In very mild conditions it is shown that, for strongly convex functions, the algorithm is global convergent. Computational results and performance profiles for a set consisting of 700 unconstrained optimization problems show that this new scaled nonlinear conjugate gradient algorithm substantially outperforms known conjugate gradient methods including: the spectral conjugate gradient SCG by Birgin and Martínez, the scaled Fletcher and Reeves, the Polak and Ribière algorithms and the CONMIN by (Shanno, D.F. and Phua, K.H., 1976, Algorithm 500, Minimization of unconstrained multivariate functions. ACM Transactions on Mathematical Software, 2, 87–94).     

Another hybrid conjugate gradient algorithm for unconstrained optimization

Another hybrid conjugate gradient algorithm is subject to analysis. The parameter β _k is computed as a convex combination of (Hestenes-Stiefel) and (Dai-Yuan) algorithms, i.e. . The parameter θ _k in the convex combination is computed in such a way so that the direction corresponding to the conjugate gradient algorithm to be the Newton direction and the pair (s _k , y _k ) to satisfy the quasi-Newton equation , where and . The algorithm uses the standard Wolfe line search conditions. Numerical comparisons with conjugate gradient algorithms show that this hybrid computational scheme outperforms the Hestenes-Stiefel and the Dai-Yuan conjugate gradient algorithms as well as the hybrid conjugate gradient algorithms of Dai and Yuan. A set of 750 unconstrained optimization problems are used, some of them from the CUTE library.      

Enhancement of membrane computing model implementation on GPU by introducing matrix representation for balancing occupancy and reducing inter-block communications

In previous studies, objects of each membrane were assigned to threads of one thread block of the graphic processing unit (GPU). The number of active threads was low if the number of objects inside a membrane was low. This study represents objects of membranes as entities of a matrix. Then a sub-matrix represents the appropriate number of objects assigned to threads of each thread block to balance the load and keep the occupancy high even when the number of objects per membrane is low. The size of the sub-matrix or the appropriate number of active threads is determined automatically. Furthermore, by this approach it is possible to assign more than one membrane to each thread block and to execute communication between membranes in the same thread block without the need for time-consuming inter-block communication. For example, using the previous algorithm, for two objects per membrane the speed up is 0.6×, while for the proposed algorithm the speed up is 32.4×.     

An improved three-term conjugate gradient algorithm for solving unconstrained optimization problems

In this article, we present an improved three-term conjugate gradient algorithm for large-scale unconstrained optimization. The search directions in the developed algorithm are proved to satisfy an approximate secant equation as well as the Dai-Liao’s conjugacy condition. With the standard Wolfe line search and the restart strategy, global convergence of the algorithm is established under mild conditions. By implementing the algorithm to solve 75 benchmark test problems with dimensions from 1000 to 10,000, the obtained numerical results indicate that the algorithm outperforms the state-of-the-art algorithms available in the literature. It costs less CPU time and smaller number of iterations in solving the large-scale unconstrained optimization.     

Parallel Resampling in the Particle Filter

Modern parallel computing devices, such as the graphics processing unit (GPU), have gained significant traction in scientific and statistical computing. They are particularly well-suited to data-parallel algorithms such as the particle filter, or more generally sequential Monte Carlo (SMC), which are increasingly used in statistical inference. SMC methods carry a set of weighted particles through repeated propagation, weighting, and resampling steps. The propagation and weighting steps are straightforward to parallelize, as they require only independent operations on each particle. The resampling step is more difficult, as standard schemes require a collective operation, such as a sum, across particle weights. Focusing on this resampling step, we analyze two alternative schemes that do not involve a collective operation (Metropolis and rejection resamplers), and compare them to standard schemes (multinomial, stratified, and systematic resamplers). We find that, in certain circumstances, the alternative resamplers can perform significantly faster on a GPU, and to a lesser extent on a CPU, than the standard approaches. Moreover, in single precision, the standard approaches are numerically biased for upward of hundreds of thousands of particles, while the alternatives are not. This is particularly important given greater single- than double-precision throughput on modern devices, and the consequent temptation to use single precision with a greater number of particles. Finally, we provide auxiliary functions useful for implementation, such as for the permutation of ancestry vectors to enable in-place propagation. Supplementary materials are available online.     

Convergence acceleration of preconditioned conjugate gradient solver based on error vector sampling for a sequence of linear systems

In this article, we focus on solving a sequence of linear systems that have identical (or similar) coefficient matrices. For this type of problem, we investigate subspace correction (SC) and deflation methods, which use an auxiliary matrix (subspace) to accelerate the convergence of the iterative method. In practical simulations, these acceleration methods typically work well when the range of the auxiliary matrix contains eigenspaces corresponding to small eigenvalues of the coefficient matrix. We develop a new algebraic auxiliary matrix construction method based on error vector sampling in which eigenvectors with small eigenvalues are efficiently identified in the solution process. We use the generated auxiliary matrix for convergence acceleration in the following solution step. Numerical tests confirm that both SC and deflation methods with the auxiliary matrix can accelerate the solution process of the iterative solver. Furthermore, we examine the applicability of our technique to the estimation of the condition number of the coefficient matrix. We also present the algorithm of the preconditioned conjugate gradient method with condition number estimation.     

A three-term conjugate gradient algorithm for large-scale unconstrained optimization problems

In this paper, a three-term conjugate gradient algorithm is developed for solving large-scale unconstrained optimization problems. The search direction at each iteration of the algorithm is determined by rectifying the steepest descent direction with the difference between the current iterative points and that between the gradients. It is proved that such a direction satisfies the approximate secant condition as well as the conjugacy condition. The strategies of acceleration and restart are incorporated into designing the algorithm to improve its numerical performance. Global convergence of the proposed algorithm is established under two mild assumptions. By implementing the algorithm to solve 75 benchmark test problems available in the literature, the obtained results indicate that the algorithm developed in this paper outperforms the existent similar state-of-the-art algorithms.     

The computation of bounds for the norm of the error in the conjugate gradient algorithm

In this paper we consider computing estimates of the norm of the error in the conjugate gradient (CG) algorithm. Formulas were given in a paper by Golub and Meurant (1997). Here, we first prove that these expressions are indeed upper and lower bounds for the A-norm of the error. Moreover, starting from these formulas, we investigate the computation of the l ₂-norm of the error. Finally, we define an adaptive algorithm where the approximations of the extreme eigenvalues that are needed to obtain upper bounds are computed when running CG leading to an improvement of the upper bounds for the norm of the error. Numerical experiments show the effectiveness of this algorithm. This revised version was published online in August 2006 with corrections to the Cover Date.     

Weighted graph based ordering techniques for preconditioned conjugate gradient methods

We describe the basis of a matrix ordering heuristic for improving the incomplete factorization used in preconditioned conjugate gradient techniques applied to anisotropic PDE's. Several new matrix ordering techniques, derived from well-known algorithms in combinatorial graph theory, which attempt to implement this heuristic, are described. These ordering techniques are tested against a number of matrices arising from linear anisotropic PDE's, and compared with other matrix ordering techniques. A variation of RCM is shown to generally improve the quality of incomplete factorization preconditioners.This work was supported by by the Natural Sciences and Engineering Research Council of Canada, and by the Information Technology Research Center, which is funded by the Province of Ontario.     

A scaled BFGS preconditioned conjugate gradient algorithm for unconstrained optimization

This letter presents a scaled memoryless BFGS preconditioned conjugate gradient algorithm for solving unconstrained optimization problems. The basic idea is to combine the scaled memoryless BFGS method and the preconditioning technique in the frame of the conjugate gradient method. The preconditioner, which is also a scaled memoryless BFGS matrix, is reset when the Powell restart criterion holds. The parameter scaling the gradient is selected as the spectral gradient. Computational results for a set consisting of 750 test unconstrained optimization problems show that this new scaled conjugate gradient algorithm substantially outperforms known conjugate gradient methods such as the spectral conjugate gradient SCG of Birgin and Martínez [E. Birgin, J.M. Martínez, A spectral conjugate gradient method for unconstrained optimization, Appl. Math. Optim. 43 (2001) 117–128] and the (classical) conjugate gradient of Polak and Ribière [E. Polak, G. Ribière, Note sur la convergence de méthodes de directions conjuguées, Revue Francaise Informat. Reserche Opérationnelle, 3e Année 16 (1969) 35–43], but subject to the CPU time metric it is outperformed by L-BFGS [D. Liu, J. Nocedal, On the limited memory BFGS method for large scale optimization, Math. Program. B 45 (1989) 503–528; J. Nocedal. http://www.ece.northwestern.edu/~nocedal/lbfgs.html].     

Global convergence of conjugate gradient method

In this article, we consider the global convergence of the Polak–Ribiére–Polyak conjugate gradient method (abbreviated PRP method) for minimizing functions that have Lipschitz continuous partial derivatives. A novel form of non-monotone line search is proposed to guarantee the global convergence of the PRP method. It is also shown that the PRP method has linear convergence rate under some mild conditions when the non-monotone line search reduces to a related monotone line search. The new non-monotone line search needs to estimate the Lipschitz constant of the gradients of objective functions, for which two practical estimations are proposed to help us to find a suitable initial step size for the PRP method. Numerical results show that the new line search approach is efficient in practical computation.     

Error norm estimation and stopping criteria in preconditioned conjugate gradient iterations

Some techniques suitable for the control of the solution error in the preconditioned conjugate gradient method are considered and compared. The estimation can be performed both in the course of the iterations and after their termination.The importance of such techniques follows from the non‐existence of some reasonable a priori error estimate for very ill‐conditioned linear systems when sufficient information about the right‐hand side vector is lacking. Hence, some a posteriori estimates are required, which make it possible to verify the quality of the solution obtained for a prescribed right‐hand side. The performance of the considered error control procedures is demonstrated using real‐world large‐scale linear systems arising in computational mechanics. Copyright © 2001 John Wiley & Sons, Ltd.     

Understanding GPU Programming for Statistical Computation: Studies in Massively Parallel Massive Mixtures

This article describes advances in statistical computation for large-scale data analysis in structured Bayesian mixture models via graphics processing unit (GPU) programming. The developments are partly motivated by computational challenges arising in fitting models of increasing heterogeneity to increasingly large datasets. An example context concerns common biological studies using high-throughput technologies generating many, very large datasets and requiring increasingly high-dimensional mixture models with large numbers of mixture components. We outline important strategies and processes for GPU computation in Bayesian simulation and optimization approaches, give examples of the benefits of GPU implementations in terms of processing speed and scale-up in ability to analyze large datasets, and provide a detailed, tutorial-style exposition that will benefit readers interested in developing GPU-based approaches in other statistical models. Novel, GPU-oriented approaches to modifying existing algorithms software design can lead to vast speed-up and, critically, enable statistical analyses that presently will not be performed due to compute time limitations in traditional computational environments. Supplemental materials are provided with all source code, example data, and details that will enable readers to implement and explore the GPU approach in this mixture modeling context.     

A parallel version of the preconditioned conjugate gradient method for boundary element equations

The parallel version of precondition techniques is developed for matrices arising from the Galerkin boundary element method for two-dimensional domains with Dirichlet boundary conditions. Results were obtained for implementations on a transputer network as well as on an nCUBE-2 parallel computer showing that iterative solution methods are very well suited for a MIMD computer. A comparison of numerical results for iterative and direct solution methods is presented and underlines the superiority of iterative methods for large systems.     

带参数共轭梯度法簇的全局收敛性

共轭梯度法是最优化中最常用的方法之一,广泛地应用于求解大规模优化问题,其中参数β_k的不同选取可以构成不同的共轭梯度法.给出了一类含有三个参数的共轭梯度算法,这种算法能够在给定的条件下证明选定的β_k在每一步都能产生一个下降方向,同时在强Wolfe线搜索下,这种算法具有全局收敛性.