Slow Convergence (slow + convergence)

Distribution by Scientific Domains


Selected Abstracts


Video completion and synthesis

COMPUTER ANIMATION AND VIRTUAL WORLDS (PREV: JNL OF VISUALISATION & COMPUTER ANIMATION), Issue 3-4 2008
Chunxia Xiao
Abstract This paper presents a new exemplar-based framework for video completion, allowing aesthetically pleasing completion of large space-time holes. We regard video completion as a discrete global optimization on a 3D graph embedded in the space-time video volume. We introduce a new objective function which enforces global spatio-temporal consistency among patches that fill the hole and surrounding it, in terms of both color similarity and motion similarity. The optimization is solved by a novel algorithm, called weighted priority belief propagation (BP), which alleviates the problems of slow convergence and intolerable storage size when using the standard BP. This objective function can also handle video texture synthesis by extending an input video texture to a larger texture region. Experiments on a wide variety of video examples with complex dynamic scenes demonstrate the advantages of our method over existing techniques: salient structures and motion information are much better restored. Copyright © 2008 John Wiley & Sons, Ltd. [source]


Initialization Strategies in Simulation-Based SFE Eigenvalue Analysis

COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING, Issue 5 2005
Song Du
Poor initializations often result in slow convergence, and in certain instances may lead to an incorrect or irrelevant answer. The problem of selecting an appropriate starting vector becomes even more complicated when the structure involved is characterized by properties that are random in nature. Here, a good initialization for one sample could be poor for another sample. Thus, the proper eigenvector initialization for uncertainty analysis involving Monte Carlo simulations is essential for efficient random eigenvalue analysis. Most simulation procedures to date have been sequential in nature, that is, a random vector to describe the structural system is simulated, a FE analysis is conducted, the response quantities are identified by post-processing, and the process is repeated until the standard error in the response of interest is within desired limits. A different approach is to generate all the sample (random) structures prior to performing any FE analysis, sequentially rank order them according to some appropriate measure of distance between the realizations, and perform the FE analyses in similar rank order, using the results from the previous analysis as the initialization for the current analysis. The sample structures may also be ordered into a tree-type data structure, where each node represents a random sample, the traverse of the tree starts from the root of the tree until every node in the tree is visited exactly once. This approach differs from the sequential ordering approach in that it uses the solution of the "closest" node to initialize the iterative solver. The computational efficiencies that result from such orderings (at a modest expense of additional data storage) are demonstrated through a stability analysis of a system with closely spaced buckling loads and the modal analysis of a simply supported beam. [source]


Multigrid convergence of inviscid fixed- and rotary-wing flows

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN FLUIDS, Issue 2 2002
C. B. Allen
Abstract The affect of multigrid acceleration implemented within an upwind-biased Euler method is presented, and applied to fixed-wing and rotary-wing flows. The convergence of fixed- and rotary-wing computations is shown to be vastly different, and multigrid is shown to be less effective for rotary-wing flows. The flow about a hovering rotor suffers from very slow convergence of the inner blade region, where the flow is effectively incompressible. Furthermore, the vortical wake must develop over several turns before convergence is achieved, whereas for fixed-wing computations the far-field grid and solution have little significance. Results are presented for single mesh and two, three, four, and five level multigrid, and using five levels a reduction in required CPU time of over 80 per cent is demonstrated for rotary-wing computations, but 94 per cent for fixed-wing computations. It is found that a simple V-cycle is the most effective, smoothing in the decreasing mesh density direction only, with a relaxed trilinear prolongation operator. Copyright © 2002 John Wiley & Sons, Ltd. [source]


Characterizing arbitrarily slow convergence in the method of alternating projections

INTERNATIONAL TRANSACTIONS IN OPERATIONAL RESEARCH, Issue 4 2009
Heinz H. Bauschke
Abstract Bauschke, Borwein, and Lewis have stated a trichotomy theorem that characterizes when the convergence of the method of alternating projections can be arbitrarily slow. However, there are two errors in their proof of this theorem. In this note, we show that although one of the errors is critical, the theorem itself is correct. We give a different proof that uses the multiplicative form of the spectral theorem, and the theorem holds in any real or complex Hilbert space, not just in a real Hilbert space. [source]


Learning scheduling control knowledge through reinforcements

INTERNATIONAL TRANSACTIONS IN OPERATIONAL RESEARCH, Issue 2 2000
K. Miyashita
Abstract This paper introduces a method of learning search control knowledge in schedule optimization problems through application of reinforcement learning. Reinforcement learning is an effective approach for the problem faced by the agent that learns its behavior through trial-and-error interactions with a dynamic environment. Nevertheless, reinforcement learning has a difficulty of slow convergence when applied to the problems with a large state space. The paper discusses the case-based function approximation technique, which makes reinforcement learning applicable to the large scale problems such as a job-shop scheduling problem. To show effectiveness of the approach, reinforcement learning is applied to acquire search control knowledge in repair-based schedule optimization process. Preliminary experiment results show that repair-action selection made by learned search control knowledge succeeded in improving scheduling quality efficiently. [source]


Further Evidence on PPP Adjustment Speeds: the Case of Effective Real Exchange Rates and the EMS,

OXFORD BULLETIN OF ECONOMICS & STATISTICS, Issue 4 2003
Ivan Paya
Abstract Two different approaches intend to resolve the ,puzzling' slow convergence to purchasing power parity (PPP) reported in the literature [see Rogoff (1996), Journal of Economic Literature, Vol. 34.] On the one hand, there are models that consider a non-linear adjustment of real exchange rate to PPP induced by transaction costs. Such costs imply the presence of a certain transaction band where adjustment is too costly to be undertaken. On the other hand, there are models that relax the ,classical' PPP assumption of constant equilibrium real exchange rates. A prominent theory put together by Balassa (1964, Journal of Political Economy, Vol. 72) and Samuelson (1964 Review of Economics and Statistics, Vol. 46), the BS effect, suggests that a non-constant real exchange rate equilibrium is induced by different productivity growth rates between countries. This paper reconciles those two approaches by considering an exponential smooth transition-in-deviation non-linear adjustment mechanism towards non-constant equilibrium real exchange rates within the EMS (European Monetary System) and effective rates. The equilibrium is proxied, in a theoretically appealing manner, using deterministic trends and the relative price of non-tradables to proxy for BS effects. The empirical results provide further support for the hypothesis that real exchange rates are well described by symmetric, nonlinear processes. Furthermore, the half-life of shocks in such models is found to be dramatically shorter than that obtained in linear models. [source]


Univariate and multirater ordinal cumulative link regression with covariate specific cutpoints

THE CANADIAN JOURNAL OF STATISTICS, Issue 4 2000
Hemant Ishwaran
Abstract The author considers a reparameterized version of the Bayesian ordinal cumulative link regression model as a tool for exploring relationships between covariates and "cutpoint" parameters. The use of this parameterization allows one to fit models using the leapfrog hybrid Monte Carlo method, and to bypass latent variable data augmentation and the slow convergence of the cutpoints which it usually entails. The proposed Gibbs sampler is not model specific and can be easily modified to handle different link functions. The approach is illustrated by considering data from a pediatric radiology study. RÉSUMÉ L'auteur propose une nouvelle paramé'trisation du modèle de régression ordinale bayésien à lien cumu-latif dont il se sert pour explorer la relation entre des covariables et des "points de coupure." Cette reparamétrisation permet d'ajuster les modèles par une méthode de Monte-Carlo à saute-mouton modifiée, évitant ainsi le besoin d'augmentation de données de la variable latente et la lenteur de convergence des points de coupure qui en découle souvent. L'échantillonneur de Gibbs qui est proposé n'est pas spécifique au modèle et peut ,tre adapté facilement à d'autres fonctions de lien. La méthode est illustrée au moyen d'une étude de radiologie pédiatrique [source]