In this paper a family of adaptive algorithms robust to impulsive noise and with low computational cost are presented. Unlike other approaches, no cost functions or filtering of the gradient are considered in order to...
详细信息
ISBN:
(纸本)9783902661692
In this paper a family of adaptive algorithms robust to impulsive noise and with low computational cost are presented. Unlike other approaches, no cost functions or filtering of the gradient are considered in order to update the filter coefficients. Its initial basis is the basic LMS algorithm and its sign-error variant. The proposed algorithms can be considered as some sign-error variants of the LMS algorithm. The algorithms are successfully tested in terms of accuracy and convergence in a standard system identification simulation in which an impulsive noise is present. Simulations show that they improve the performance of LMS variants that are robust to impulsive noise.
Two research subjects in geosciences which lately underwent significant progress are treated in this review. In the first part, we focus on one key ingredient for the numerical approximation of the Darcy flow problem,...
详细信息
Two research subjects in geosciences which lately underwent significant progress are treated in this review. In the first part, we focus on one key ingredient for the numerical approximation of the Darcy flow problem, namely the discretization of diffusion terms on general polygonal/polyhedral meshes. We present different schemes and discuss in detail their fundamental numerical properties such as stability, consistency, and robustness. The second part of the paper is devoted to error control and adaptivity for model problems in geosciences. We present the available a posteriori estimates guaranteeing the maximal overall error and show how the different error components can be identified. These estimates are used to formulate adaptive stopping criteria for linear and nonlinear solvers, time step choice adjustment, and adaptive mesh refinement. Numerical experiments illustrate such entirely adaptive algorithms.
The aim of this paper is to develop efficient online adaptive algorithms for the generalized eigen-decomposition problem which arises in a variety of modern signal processing applications. First, we reinterpret the ge...
详细信息
The aim of this paper is to develop efficient online adaptive algorithms for the generalized eigen-decomposition problem which arises in a variety of modern signal processing applications. First, we reinterpret the generalized eigen-decomposition problem as an unconstrained minimization problem by constructing a novel cost function. Second, by applying projection approximation method and recursive least-square (RLS) technique to the cost function, a parallel adaptive algorithm for a basis for the r-dimensional (r > 0) dominant generalized eigen-subspace and a sequential algorithm based on deflation technique for the first r-dominant generalized eigenvectors are derived. These algorithms can be viewed as counterparts of the extended projection approximation subspace tracking (PAST) and PASTd algorithms, respectively. Furthermore, we modify the parallel algorithm to explicitly estimate the first r-generalized eigenvectors in parallel, not the generalized eigen-subspace. More important, the modified parallel algorithm can be used to extract multiple generalized eigenvectors of two nonstationary sequences, while the proposed sequential algorithm lacks this ability because of slow convergence of minor generalized eigenvectors due to error propagation of the deflation technique. Third, following convergence analysis methods for PAST and PASTd, we prove the asymptotic convergence properties of the proposed algorithms. Finally, computer simulations are performed to investigate the accuracy and the speed advantages of the proposed algorithms.
In this article, we introduce accelerated algorithms for. linear discriminant analysis (LDA) and feature extraction from unimodal multiclass Gaussian data. Current adaptive methods based on the gradient descent optimi...
详细信息
In this article, we introduce accelerated algorithms for. linear discriminant analysis (LDA) and feature extraction from unimodal multiclass Gaussian data. Current adaptive methods based on the gradient descent optimization technique use a fixed or a monotonically decreasing step size in each iteration, which results in a slow convergence rate. Here, we use a variable step size, optimally computed in each iteration using the steepest descent method, in order to accelerate the convergence of the algorithm. Based on the new adaptive algorithm, we present a self-organizing neural network for adaptive computation of the square root of the inverse covariance matrix (Sigma(-1/2)) and use it (i) in a network for optimal feature extraction from Gaussian data and (ii) in cascaded form with a principal component analysis network for LDA. Experimental results demonstrate fast convergence and high stability of the algorithm and justify its advantages for on-line pattern recognition applications with stationary and non-stationary input data. (C) 2003 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.
Structural properties are examined of systems with physical component values as parameters. Both state variable realizations and transfer function descriptions are investigated. The transfer functions in particular ar...
详细信息
Structural properties are examined of systems with physical component values as parameters. Both state variable realizations and transfer function descriptions are investigated. The transfer functions in particular are shown to be the ratios of polynomials with coefficients multilinear in the parameters. These structures prove useful in formulating adaptive algorithms.
The aim in blind source separation is to separate linear mixtures of statistically independent non-Gaussian signals without resorting to an a priori knowledge of the sources or the mixing system. In this paper we prop...
详细信息
The aim in blind source separation is to separate linear mixtures of statistically independent non-Gaussian signals without resorting to an a priori knowledge of the sources or the mixing system. In this paper we propose a new family of adaptive algorithms that recursively compute the optimum separating system. The algorithms are of the gradient ascent type and maximize a statistical criterion that involves only second- and fourth-order cumulants. We present a complete analysis of all the stationary points in the proposed criterion for an arbitrary number of complex sources. We demonstrate that the algorithms can only converge to points where perfect separation is achieved provided that the mixing system is a square invertible matrix and all the sources have the same kurtosis sign. We also prove that the criterion is free of undesirable maxima. (C) 1999 Elsevier Science B.V. All rights reserved.
This study considers a commutation error (CE) that results from a difference associated with the altered sequence in real active noise control (ANC) applications as compared with that at the derivation stage. New adap...
详细信息
This study considers a commutation error (CE) that results from a difference associated with the altered sequence in real active noise control (ANC) applications as compared with that at the derivation stage. New adaptive algorithms are developed as FxLMS/CE, FxNLMS/CE and FxRLS/CE in an aim to eliminate the CE-associated disturbance and to liberate the restriction of slow adaptation imposed on the existing adaptive algorithms in the ANC applications. Computer simulations show that the rate of convergence is greatly improved for the new adaptive algorithms as compared with that of the conventional algorithms. In parallel with the improved rate of convergence, simulations exhibit efficient ANC performance for all CE-based algorithms. The best ANC performance is seen for FxRLS/CE algorithm that can acquire similar to 2 s of convergence rate and similar to 34 dB reduction of sound pressure level for band-limited white noise. All experimental results indeed demonstrate enhanced ANC performance;the FxNLMS/CE algorithm can acquire similar to 2 s of convergence rate and similar to 20 dB reduction of sound pressure level for band-limited white noise. Our data together support the effectiveness to include CE into the FIR filter-based adaptive algorithms for superior ANC performance with respect to the convergence speed and noise reduction level. (C) 2006 Elsevier Ltd. All rights reserved.
This paper presents a coordinated control of electronic stability control (ESC) and active front steering (AFS) with adaptive algorithms for yaw moment distribution in integrated chassis control (ICC). In order to dis...
详细信息
This paper presents a coordinated control of electronic stability control (ESC) and active front steering (AFS) with adaptive algorithms for yaw moment distribution in integrated chassis control (ICC). In order to distribute a control yaw moment into control tire forcres of ESC and AFS, and to coordinate the relative usage of ESC to AFS, a LMS/Newton algorithm (LMSN) is adopted. To make the control tire forces zero in applying LMS and LMSN, the zero-attracting mechanism is adopted. Simulations on vehicle simulation software, CarSimA (R), show that the proposed algorithm is effective for yaw moment distribution in integrated chassis control.
This special issue of Journal of Computational Science follows the Agent-Based Simulations, adaptive algorithms and Solvers (ABS-AAS) Workshop in frame of the International Conference on Computational Science (ICCS) h...
详细信息
This special issue of Journal of Computational Science follows the Agent-Based Simulations, adaptive algorithms and Solvers (ABS-AAS) Workshop in frame of the International Conference on Computational Science (ICCS) held in Reykjavik, Iceland, in June 1-3, 2015. The aim of this workshop was to integrate results of different domains of computer science, computational science and mathematics. Chairmans of the ABS-AAS Worksop invited papers oriented toward simulations, either hard simulations by means of finite element or finite difference methods, or soft simulations by means of agent-based systems, evolutionary computations, and other. This was thirteen ABS-AAS workshop in frame of the ICCS conference. The workshop was organized by five co-chairmens, including Maciej Paszynski, Robert Schaefer and Krzysztof Cetnarowicz from AGH University, Krakow, Poland, David Pardo from the University of the Basque Country, Bilbao, Spain and Victor Calo from King Abdullah University of Science and Technology. (C) 2015 Elsevier B.V. All rights reserved.
The dichotomous coordinate descent (DCD) algorithm has been successfully used for significant reduction in the complexity of recursive least squares (RLS) algorithms. In this brief, we generalize the application of th...
详细信息
The dichotomous coordinate descent (DCD) algorithm has been successfully used for significant reduction in the complexity of recursive least squares (RLS) algorithms. In this brief, we generalize the application of the DCD algorithm to RLS adaptive filtering in impulsive noise scenarios and derive a unified update formula. By employing different robust strategies against impulsive noise, we develop novel computationally efficient DCD-based robust recursive algorithms. Furthermore, to equip the proposed algorithms with the ability to track abrupt changes in unknown systems, a simple variable forgetting factor mechanism is also developed. Simulation results for channel identification scenarios in impulsive noise demonstrate the effectiveness of the proposed algorithms.
暂无评论