This paper focuses on a locally recurrent multilayer network with internal feedback paths, the IIR-MLP. The computation of the partial derivatives of the network's output with respect to its trainable weights is a...
详细信息
This paper focuses on a locally recurrent multilayer network with internal feedback paths, the IIR-MLP. The computation of the partial derivatives of the network's output with respect to its trainable weights is achieved using backpropagation through adjoints and a second order global recursive prediction error (GRPE) training algorithm is developed. Also, a local version of the GRPE is presented in order to cope with the increased computational burden of the global version. The efficiency of the proposed learning schemes, as compared to conventional gradient based methods, is tested on the wind prediction problem from 15 min to 3 h ahead on a site, using spatial correlation and facilitating measurements from nearby sites up to 40 km away.
In this paper, two new correntropy-based data-selective algorithms using the prediction error method and proportionate principle are proposed for acoustic feedback cancellation (AFC). The proposed data selective appro...
详细信息
ISBN:
(数字)9798350362510
ISBN:
(纸本)9798350362527
In this paper, two new correntropy-based data-selective algorithms using the prediction error method and proportionate principle are proposed for acoustic feedback cancellation (AFC). The proposed data selective approach is applied for the improved practical variable step size proportionate normalized least mean square algorithm (IPNLMS-IPVSS) and its novel tanh-based version. It is shown that the proposed data-selective algorithms can achieve close performance to the original algorithms at a reduced average numerical complexity.
Suppose L simultaneous independent stochastic systems generate observations, where the observations from each system depend on the underlying parameter of that system. The observations are unlabeled (anonymized), in t...
详细信息
Suppose L simultaneous independent stochastic systems generate observations, where the observations from each system depend on the underlying parameter of that system. The observations are unlabeled (anonymized), in the sense that an analyst does not know which observation came from which stochastic system. How can the analyst estimate the underlying parameters of the L systems? Since the anonymized observations at each time are an unordered set of L measurements (rather than a vector), classical stochastic gradient algorithms cannot be directly used. By using symmetric polynomials, we formulate a symmetric measurement equation that maps the observation set to a unique vector. We then construct an adaptive filtering algorithm that yields a statistically consistent estimate of the underlying parameters.
In this paper, distributed fusion filtering problem in multisensory dynamic system is considered. The approximation scheme for calculation of cross-covariances is presented, which are based on a correlation coefficien...
详细信息
ISBN:
(纸本)9781424455690
In this paper, distributed fusion filtering problem in multisensory dynamic system is considered. The approximation scheme for calculation of cross-covariances is presented, which are based on a correlation coefficient in steady-state. In addition, a new limiting cross-covariance method is also proposed, which gives the opportunity to save calculation times, and is computationally useful to implement the fusion filters in a real-time application. Numerical examples demonstrating the effectiveness of the proposed algorithm are presented.
This paper develops algorithms for filtering and smoothing for parallel computers. Numerical results are presented and implementation details are discussed. In the example it is illustrated that parallel methods have ...
详细信息
This paper develops algorithms for filtering and smoothing for parallel computers. Numerical results are presented and implementation details are discussed. In the example it is illustrated that parallel methods have better convergence properties than nonparallel methods for nonlinear problems.
This paper develops information filter and array algorithms for the linear minimum mean square error estimator (LMMSE) for discrete-time Markovian jump linear systems (MJLSs). A numerical example to show the advantage...
详细信息
This paper develops information filter and array algorithms for the linear minimum mean square error estimator (LMMSE) for discrete-time Markovian jump linear systems (MJLSs). A numerical example to show the advantage of the array algorithm in fix-point implementations is presented in order to demonstrate the advantage of this approach.
In this paper, we describe a new parameter estimation methodology called transpose domain filtering (TDF). TDF is a processing formulation that that is different from conventional processing (e.g., beamscan processing...
详细信息
In this paper, we describe a new parameter estimation methodology called transpose domain filtering (TDF). TDF is a processing formulation that that is different from conventional processing (e.g., beamscan processing) in that the weight vector is applied in the domain that lies transpose to the signal model. For a typical radar array processing problem, this means TDF would apply the weight vector in the snapshot domain as opposed to the element domain. We provide two examples of TDF applied to minimum variance distortionless response (MVDR) beam scan and code division multiple access (CDMA) synchronization. We later show that rank reduction techniques can be used in TDF, and then prove that when principal components rank reduction is applied to TDF, the result is mathematically equivalent to the better-known multiple signal characterization (MUSIC) algorithm.
Recommender systems provide personalized recommendations for products, information or services in electronic commerce. Collaborative filtering is one of the most successful techniques that attempts to recommend items ...
详细信息
Recommender systems provide personalized recommendations for products, information or services in electronic commerce. Collaborative filtering is one of the most successful techniques that attempts to recommend items (such as music, movies, web sites) which are likely to be interested to some people. In this paper we study collaborative filtering algorithms from the influences of users and the topology of network. We analyze relationships between users from their scores on same items and how influential users impact others. Whether an item is recommended to a user, we not choose nearest neighbors from all users, but choose them from the influential users for the user, so the neighbors are nearest and strongest influential than others for the user, making recommendation results are more effective and efficiency, which is the principle of the improved algorithm. Relationships between users construct a complex network, so attributes and properties of complex network are also combined with the algorithm of collaborative filtering. Numerical and contrast experiments are proposed to explain and prove the feasibility and efficiency of the recommendation algorithm of collaborative filtering based on influence and complex network. The experimental results show that efficiency of the improved algorithm is better than traditional collaborative filtering algorithm.
Speckle filtering in multiresolution frameworks is the key to extend spatial adaptivity also across scales. Several procedures based on wavelets and pyramids, in which the different resolutions are adaptively filtered...
详细信息
Speckle filtering in multiresolution frameworks is the key to extend spatial adaptivity also across scales. Several procedures based on wavelets and pyramids, in which the different resolutions are adaptively filtered, have been introduced for selective speckle smoothing with preservation of features and point targets. In this work, comparison tests on true and synthetic speckled images show significant advances of the multi- over the single-scale approach, especially for the pyramid based scheme proposed by the authors, where visual comparisons on SAR images show better texture preservation.
暂无评论