We propose a generalization of the Weierstrass iteration for over-constrained systems of equations and we prove that the proposed method is the Gauss-Newton iteration to find the nearest system which has at least k common roots and which is obtained via a perturbation of prescribed structure. In the univariate case we show the connection of our method to the optimization problem formulated by K...

We present a horizon-based value iteration algorithm called Reverse Value Iteration (RVI). Empirical results on a variety of domains, both synthetic and real, show RVI often yields speedups of several orders of magnitude. RVI does this by ordering backups by horizons, with preference given to closer horizons, thereby avoiding many unnecessary and incorrect backups. We also compare to related wo...

ABSTRACT— Digital signal processing algorithms are described by iterative data-flow graphs where nodes represent computations and edges represent communications. In this paper we propose a novel method to determine the iteration bound, which is the fundamental lower bound of the iteration period of a processing algorithm, by using the minimum cycle mean algorithm to achieve a lower polynomial t...

We propose an accelerating method for the restarted Arnoldi iteration to compute a number of eigenvalues of the standard eigenproblem Ax = x and discuss the dependence of the convergence rate of the accelerated iteration on the distribution of spectrum. The e ectiveness of the approach is proved by numerical results. We also propose a new parallelization technique for the nonsymmetric double sh...

We propose and analyze a new parallel coordinate descent method—‘NSync— in which at each iteration a random subset of coordinates is updated, in parallel, allowing for the subsets to be chosen non-uniformly. We derive convergence rates under a strong convexity assumption, and comment on how to assign probabilities to the sets to optimize the bound. The complexity and practical performance of th...

A new algorithm for solving linear complementarity problems with suucient matrices is proposed. If the problem has a solution the algorithm is superlinearly convergent from any positive starting points, even for degenerate problems. Each iteration requires only one matrix factorization and at most two backsolves. Only one backsolve is necessary if the problem is known to be nondegenerate. The a...

We present an approach for the acceleration of the restarted Arnoldi iteration for the computation of a number of eigenvalues of the standard eigenproblem Ax = λx. This study applies the Chebyshev polynomial to the restarted Arnoldi iteration and proves that it computes necessary eigenvalues with far less complexity than the QR method. We also discuss the dependence of the convergence rate of t...

This paper presents a two-iteration concatenated Bose-Chaudhuri-Hocquenghem (BCH) code and its highspeed low-complexity two-parallel decoder architecture for 100 Gb/s optical communications. The proposed architecture features a very high data processing rate as well as excellent error correction capability. A low-complexity syndrome computation architecture and a high-speed dual-processing pipe...

A large-step infeasible-interior-point method is proposed for solving P∗(κ)-matrix linear complementarity problems. It is new even for monotone LCP. The algorithm generates points in a large neighborhood of an infeasible central path. Each iteration requires only one matrix factorization. If the problem is solvable, then the algorithm converges from arbitrary positive starting points. The compu...

The complexity of algorithms for solving Markov Decision Processes (MDPs) with finite state and action spaces has seen renewed interest in recent years. New strongly polynomial bounds have been obtained for some classical algorithms, while others have been shown to have worst case exponential complexity. In addition, new strongly polynomial algorithms have been developed. We survey these result...