Computational error-analysis of a discontinuous Galerkin discretization applied to large-eddy simulation of homogeneous turbulence

A computational error-assessment of large-eddy simulation (LES) in combination with a discontinuous Galerkin finite element method is presented for homogeneous, isotropic, decaying turbulence. The error-landscape database approach is used to quantify the total simulation error that arises from the use of the Smagorinsky eddy-viscosity model in combination with the Galerkin discretization. We adopt a modified HLLC flux, allowing an explicit control over the dissipative component of the numerical flux. The optimal dependence of the Smagorinsky parameter on the spatial resolution is determined for second and third order accurate Galerkin methods. In particular, the role of the numerical dissipation relative to the contribution from the Smagorinsky dissipation is investigated. We observed an ‘exchange of dissipation’ principle in the sense that an increased numerical dissipation implied a reduction in the optimal Smagorinsky parameter. The predictions based on Galerkin discretization with fully stabilized HLLC flux were found to be less accurate than when a central discretization with (mainly) Smagorinsky dissipation was used. This was observed for both the second and third order Galerkin discretization, suggesting to emphasize central discretization of the convective nonlinearity and stabilization that mimics eddy-viscosity as sub-filter dissipation.     

Effect of data discretization on the classification accuracy in a high‐dimensional framework

We investigate discretization of continuous variables for classification problems in a high‐ dimensional framework. As the goal of classification is to correctly predict a class membership of an observation, we suggest a discretization method that optimizes the discretization procedure using the misclassification probability as a measure of the classification accuracy. Our method is compared to several other discretization methods as well as result for continuous data. To compare performance we consider three supervised classification methods, and to capture the effect of high dimensionality we investigate a number of feature variables for a fixed number of observations. Since discretization is a data transformation procedure, we also investigate how the dependence structure is affected by this. Our method performs well, and lower misclassification can be obtained in a high‐dimensional framework for both simulated and real data if the continuous feature variables are first discretized. The dependence structure is well maintained for some discretization methods. © 2012 Wiley Periodicals, Inc.     

UniDis: a universal discretization technique

Discretization techniques have played an important role in machine learning and data mining as most methods in such areas require that the training data set contains only discrete attributes. Data discretization unification (DDU), one of the state-of-the-art discretization techniques, trades off classification errors and the number of discretized intervals, and unifies existing discretization criteria. However, it suffers from two deficiencies. First, the efficiency of DDU is very low as it conducts a large number of parameters to search good results, which does not still guarantee to obtain an optimal solution. Second, DDU does not take into account the number of inconsistent records produced by discretization, which leads to unnecessary information loss. To overcome the above deficiencies, this paper presents a Uni versal Dis cretization technique, namely UniDis. We first develop a non-parametric normalized discretization criteria which avoids the effect of relatively large difference between classification errors and the number of discretized intervals on discretization results. In addition, we define a new entropy-based measure of inconsistency for multi-dimensional variables to effectively control information loss while producing a concise summarization of continuous variables. Finally, we propose a heuristic algorithm to guarantee better discretization based on the non-parametric normalized discretization criteria and the entropy-based inconsistency. Besides theoretical analysis, experimental results demonstrate that our approach is statistically comparable to DDU evaluated by a popular statistical test and it yields a better discretization scheme which significantly improves the accuracy of classification than previously other known discretization methods except for DDU by running J4.8 decision tree and Naive Bayes classifier.     

信息表的闭离散化方案研究

提出对象域U的有序划分概念,讨论一种特殊的离散化方案（闭离散化方案）。给出对象域U的有序划分对应的闭离散化方案获取算法CDA,分析闭离散化方案与对象域U的有序划分之间的关系,证明了闭离散化方案在离散格到划分格的映射f下能保持交并运算。     

一类齐次离散双线性系统可控性分析

在系统分析中,可控性是系统的一个重要特性.在工程实际操作中,往往需要对一个连续系统进行离散化处理,人们希望系统在离散化后能保留原系统的重要系统特征,比如可控性.对于线性系统,我们有成熟的判断方法.然而,对于非线性系统则无统一的判别方法.Elliott在2005年给出了一个二阶双线性系统经过离散化后,可控性发生变化的例子.它表明一个系统在离散化前后,它的可控性可能会发生改变.本文旨在给出一类二阶离散化双线性系统可控性充分条件,并和已有结果作比较,表明本文结果更具有一般性.另外,本文对于3阶及以上的这类系统可控性做出了不可控的判断.     

Properties of a Level Set Algorithm for the Visibility Problems

We study an implicit visibility formulation and show that the corresponding closed form formula satisfies a dynamic programming principle, and is the viscosity solution of a Hamilton-Jacobi type equation involving jump discontinuities in the Hamiltonian. We derive the corresponding discretization in multi-dimensions and prove convergence of the corresponding numerical approximations. Finally, we introduce a generalization of the original Hamilton-Jacobi equation and the corresponding discretization that can be solved efficiently using either the fast sweeping or the fast marching methods. Thus, the visibility of an observer in non-constant media can be computed. We also introduce a specialization of the algorithms for environments in which occluders are described by the graph of a function.     

On the convergence of a new time window discretization method for the traveling salesman problem with time window constraints

In this paper, we discuss convergence of a time window discretization method for the traveling salesman problem with time window constraints. This recently proposed discretization method leads to a lower bound to the minimization problem. Analysis shows that this method always guarantees convergence to the optimal solution. We also illustrate that the traditional method that gives feasible solutions does not always provide such a guarantee.     

粗糙集理论中的离散化问题

一引言数据分析及数据挖掘,是一个重要的正在迅速发展的研究课题。波兰科学家Z.Pawlak于1982年提出的粗糙集(Rough Set)理论正是解决这一问题的新理论,它可用于处理决策信息表中的不确定知识,并用规则的形式表达,是一种有效的知识获取工具。在Roughset中,数据约简是非常重要的,包括属性约简和值约简。运用粗糙集理论处理决策表时,要求决策表中各值     

Binding energy of the very early universe: Abandoning einstein for a discretized three-torus poset. A proposal on the origin of dark energy

We formulate seeds of a radical theory. We propose a model discretization of the Universe based on a T ³-torus. A suggestion for the discretization of the matter Lagrangian is also given. An attempt ismade to relate the natural Planck units to the parameters of elementary geometric cells. Then, by introducing a many-body method, we speculate on the binding energy in the very early Universe. Finally, we introduce postulates of the model and some effects which we expect from it.     

Asymptotic properties of the space–time adaptive numerical solution of a nonlinear heat equation

We consider the fully adaptive space–time discretization of a class of nonlinear heat equations by Rothe’s method. Space discretization is based on adaptive polynomial collocation which relies on equidistribution of the defect of the numerical solution, and the time propagation is realized by an adaptive backward Euler scheme. From the known scaling laws, we infer theoretically the optimal grids implying error equidistribution, and verify that our adaptive procedure closely approaches these optimal grids.     

Pricing American bond options using a penalty method

We develop a novel numerical method to price American options on a discount bond under the Cox–Ingrosll–Ross (CIR) model which is governed by a partial differential complementarity problem. We first propose a penalty approach to this complementarity problem, resulting in a nonlinear partial differential equation (PDE). To numerically solve this nonlinear PDE, we develop a novel fitted finite volume method for the spatial discretization, coupled with a fully implicit time-stepping scheme. We show that this full discretization scheme is consistent, stable and monotone, and hence the convergence of the numerical solution to the viscosity solution of the continuous problem is guaranteed. To solve the discretized nonlinear system, we design an iterative method and prove that the method is convergent. Numerical results are presented to demonstrate the accuracy, efficiency and robustness of our methods.     

Discretization for naive-Bayes learning: managing discretization bias and variance

Quantitative attributes are usually discretized in Naive-Bayes learning. We establish simple conditions under which discretization is equivalent to use of the true probability density function during naive-Bayes learning. The use of different discretization techniques can be expected to affect the classification bias and variance of generated naive-Bayes classifiers, effects we name discretization bias and variance. We argue that by properly managing discretization bias and variance, we can effectively reduce naive-Bayes classification error. In particular, we supply insights into managing discretization bias and variance by adjusting the number of intervals and the number of training instances contained in each interval. We accordingly propose proportional discretization and fixed frequency discretization, two efficient unsupervised discretization methods that are able to effectively manage discretization bias and variance. We evaluate our new techniques against four key discretization methods for naive-Bayes classifiers. The experimental results support our theoretical analyses by showing that with statistically significant frequency, naive-Bayes classifiers trained on data discretized by our new methods are able to achieve lower classification error than those trained on data discretized by current established discretization methods.     

Solving a Boundary Value Problem by a Newton-Like Method

We present a semilocal convergence study where a Newton-like method is used to solve a boundary value problem of class M ; as an application of the previous study by means a scheme of discretization we approximate the solution of a particular problem.     

Finite-element Discretization of Static Hamilton-Jacobi Equations based on a Local Variational Principle

We propose a linear finite-element discretization of Dirichlet problems for static Hamilton–Jacobi equations on unstructured triangulations. The discretization is based on simplified localized Dirichlet problems that are solved by a local variational principle. It generalizes several approaches known in the literature and allows for a simple and transparent convergence theory. In this paper the resulting system of nonlinear equations is solved by an adaptive Gauss–Seidel iteration that is easily implemented and quite effective as a couple of numerical experiments show.Dedicated to Peter Deuflhard on the occasion of his 60th birthday     

A Petrov–Galerkin multilayer discretization to second order elliptic boundary value problems

We study in this paper a multilayer discretization of second order elliptic problems, aimed at providing reliable multilayer discretizations of shallow fluid flow problems with diffusive effects. This discretization is based upon the formulation by transposition of the equations. It is a Petrov–Galerkin discretization in which the trial functions are piecewise constant per horizontal layers, while the test functions are continuous piecewise linear, on a vertically shifted grid.We prove the well posedness and optimal error order estimates for this discretization in natural norms, based upon specific inf–sup conditions.We present some numerical tests with parallel computing of the solution based upon the multilayer structure of the discretization, for academic problems with smooth solutions, with results in full agreement with the theory developed.     

A family of space–time connecting discretization schemes with local impact detection for elastodynamic contact problems

We derive, formulate and analyze a new family of discretization schemes for elastodynamic contact problems which implicitly resolve the individual impact times for each node on the contact interface. Within our approach, information from the space discrete setting is incorporated into the time discretization by means of pointwise chosen parameters for the time discretization scheme. The members of this family can be interpreted as modified Newmark schemes, thus making them easily understandable and implementable. We prove that for certain parameter choices the algorithms are dissipative methods. Further, as our analysis and our numerical experiments show, a special solution dependent choice of parameters leads to a new space–time connecting discretization with a highly stable behavior of displacements, velocities and boundary stresses at the contact interface.     

Fully Discrete Analysis of a Discontinuous Finite Element Method for the Keller-Segel Chemotaxis Model

This paper formulates and analyzes fully discrete schemes for the two-dimensional Keller-Segel chemotaxis model. The spatial discretization of the model is based on the discontinuous Galerkin methods and the temporal discretization is based either on Forward Euler or the second order explicit total variation diminishing (TVD) Runge-Kutta methods. We consider Cartesian grids and prove fully discrete error estimates for the proposed methods. Our proof is valid for pre-blow-up times since we assume boundedness of the exact solution.     

Bayesian Clustering of Fuzzy Feature Vectors Using a Quasi-Likelihood Approach

Bayesian model-based classifiers, both unsupervised and supervised, have been studied extensively and their value and versatility have been demonstrated on a wide spectrum of applications within science and engineering. A majority of the classifiers are built on the assumption of intrinsic discreteness of the considered data features or on the discretization of them prior to the modeling. On the other hand, Gaussian mixture classifiers have also been utilized to a large extent for continuous features in the Bayesian framework. Often the primary reason for discretization in the classification context is the simplification of the analytical and numerical properties of the models. However, the discretization can be problematic due to its textit{ad hoc} nature and the decreased statistical power to detect the correct classes in the resulting procedure. We introduce an unsupervised classification approach for fuzzy feature vectors that utilizes a discrete model structure while preserving the continuous characteristics of data. This is achieved by replacing the ordinary likelihood by a binomial quasi-likelihood to yield an analytical expression for the posterior probability of a given clustering solution. The resulting model can be justified from an information-theoretic perspective. Our method is shown to yield highly accurate clusterings for challenging synthetic and empirical data sets.     

Preconditioning strategies for models of incompressible flow

We describe some new preconditioning strategies for handling the algebraic systems of equations that arise from discretization of the incompressible Navier-Stokes equations. We demonstrate how these methods adapt in a straightforward manner to decisions on implicit or explicit time discretization, explore their use on a collection of benchmark problems, and show how they relate to classical techniques such as projection methods and SIMPLE.