检索结果-内蒙古大学图书馆

Weight initialization with reference patterns

NEUROCOMPUTING 1998年第1-3期20卷 265-278页

作者： Lehtokangas, M Saarinen, J Tampere Univ Technol Signal Proc Lab FIN-33101 Tampere Finland

The problem of weight initialization in multilayer perceptron networks is considered. A new computationally simple weight initialization method based on the usage of reference patterns is presented. A reference pattern is a vector which is used to represent data points that fall in its vicinity in the data space. On one hand, the proposed method aims to set the initial weight values to be such that inputs to network nodes are within the active region (in other words, nodes are not saturated). On the other hand, the goal is to distribute the discriminant functions formed by the hidden units evenly into the input space area where training data is located. The proposed method is tested with the widely used two-spirals classification benchmark problem and channel equalization problem where several alternatives for obtaining suitable reference patterns are investigated. Also, the effect of the initialization is studied when two commonly used cost functions are used in the training phase. These are the mean square error and relative entropy cost functions. A comparison with the conventional random initialization shows that significant improvement in convergence can be achieved with the proposed method. In addition, the computational cost of the initialization was found to be negligible compared with the cost of training. (C) 1998 Elsevier Science B.V. All rights reserved.

关键词： weight initialization reference patterns multilayer perceptron network

来源：评论

学校读者我要写书评

暂无评论

Structure and X-ray powder reference patterns for hexagonal perovskite-related phases, (Sr_0.8Ca_0.2)₅Co₄O₁₂ and Sr₆Co₅O₁₅

引用

POWDER DIFFRACTION 2011年第1期26卷 22-30页

作者： Wong-Ng, W. Kaduk, J. A. Liu, G. Natl Inst Stand & Technol Div Ceram Gaithersburg MD 20899 USA Poly Crystallog Inc Naperville IL 60540 USA

Two selected members of the homologous series A(n+2)BB'O-n(3n+3) (A Sr and Ca;B and B' = Co) have been investigated for their crystal structures because of their potential applications as thermoelectric materials. A combined Rietveld refinement and spin-polarized magnetic geometry optimization technique was employed for the structural studies. Both the n=3 member, (Sr0.8Ca0.2)(5)Co4O12, and the n=4 member, Sr6Co5O15, have distorted hexagonal perovskite-related structures that possess one-dimensional cobalt oxide chains separated by alkaline-earth cations. The linear chains consist of one unit of CoO6 trigonal prism alternating with n units of CoO6 octahedra. Crystal structures and reference powder X-ray diffraction patterns of (Sr0.8Ca0.2)(5)Co4O11 [P3cl. a=9.4196(2) angstrom. c =19.9857(6) angstrom. V=825.83 angstrom(3), and D-x=5.358 g/cm(3)] and Sr6Co5O15 [R32, a=9.497 64(12) angstrom. c=12.3956(2) angstrom, V=968.34 angstrom(3), and D-x=5.455 g/cm(3)] are reported. (C) 2011 International Centre for Diffraction Data. [DOI:10.1154/1.3555294]

关键词： thermoelectric materials A(n+2)BB(n)' O3n+3 homologous series (Sr0.8Ca0.2)(5)Co4O12 Sr6Co5O15 crystal structure reference patterns

来源：评论

学校读者我要写书评

暂无评论

Visualizing reference patterns for solving memory leaks in Java

引用

CONCURRENCY-PRACTICE AND EXPERIENCE 2000年第14期12卷 1431-1454页

作者： De Pauw, W Sevitsky, G IBM Corp Thomas J Watson Res Ctr Yorktown Hts NY 10598 USA

Many Java programmers believe that they do not have to worry about memory management because of automatic garbage collection. In fact, many Java programs run out of memory unexpectedly after performing a number of operations. A memory leak in Java is caused when an object that is no longer needed cannot be reclaimed because another object is still referring to it. Memory leaks can be difficult to solve, since the complexity of most programs prevents us from manually verifying the validity of every reference. In this paper we show a new methodology for finding the causes of memory leaks. We have identified a basic memory leak scenario which fits many important cases, In this scenario, we allow the programmer to identify a period of time in which temporary objects are expected to be created and released. Using this information we are able to identify objects that persist beyond this period and the references which are holding on to them. Scaling this methodology to real-world systems brings additional challenges. We propose a novel combination of visual syntax and reference pattern extraction to manage this additional complexity. We also describe how these techniques can be applied to a wider class of memory problems, including the exploration of large data structures. These techniques have been implemented and have been proven successful on large projects, Copyright (C) 2000 John Wiley & Sons, Ltd.

关键词： memory leak detection garbage collection program visualization reference patterns data structures program analysis

来源：评论

学校读者我要写书评

暂无评论

Visualizing reference patterns for solving memory leaks in Java

引用

Concurrency and Computation: Practice and Experience 2000年第14期12卷

作者： Wim De Pauw Gary Sevitsky IBM T. J. Watson Research Center P.O. Box 704 Yorktown Heights NY 10598 U.S.A.

Many Java programmers believe that they do not have to worry about memory management because of automatic garbage collection. In fact, many Java programs run out of memory unexpectedly after performing a number of operations. A memory leak in Java is caused when an object that is no longer needed cannot be reclaimed because another object is still referring to it. Memory leaks can be difficult to solve, since the complexity of most programs prevents us from manually verifying the validity of every reference. In this paper we show a new methodology for finding the causes of memory leaks. We have identified a basic memory leak scenario which fits many important cases. In this scenario, we allow the programmer to identify a period of time in which temporary objects are expected to be created and released. Using this information we are able to identify objects that persist beyond this period and the references which are holding on to them. Scaling this methodology to real-world systems brings additional challenges. We propose a novel combination of visual syntax and reference pattern extraction to manage this additional complexity. We also describe how these techniques can be applied to a wider class of memory problems, including the exploration of large data structures. These techniques have been implemented and have been proven successful on large projects. Copyright © 2000 John Wiley & Sons, Ltd.

关键词： memory leak detection garbage collection program visualization reference patterns data structures program analysis

来源：评论

学校读者我要写书评

暂无评论

Identification and interpretation of generalized two-dimensional correlation spectroscopy features through decomposition of the perturbation domain

引用

APPLIED SPECTROSCOPY 2003年第12期57卷 1561-1574页

作者： Schulze, G Jirasek, A Blades, MW Turner, RFB Univ British Columbia Biotechnol Lab Vancouver BC V6T 1Z3 Canada Univ British Columbia Dept Chem Vancouver BC V6T 1Z1 Canada Univ British Columbia Dept Elect Engn Vancouver BC V6T 1Z4 Canada

Generalized two-dimensional correlation spectroscopy offers great scope for revealing the behavior of relationships between components of a system under empirical study. We have developed methods that aid in the interpretation of two-dimensional correlation spectroscopy. These methods include reference patterns for two-dimensional correlation and correlation coefficient maps, their superposition and joint interpretation, and the use of delta functions to decompose them in the perturbation domain. We show how their joint use permits discrimination between similar two-dimensional correlation map features on the basis of different correlation coefficients. We also show how the decomposition of maps into the perturbation domain reflects the dynamic behavior of spectral features over the course of the perturbation and permits discrimination between otherwise highly similar two-dimensional correlation crosspeaks. These approaches simplify the interpretation of two-dimensional correlation spectroscopy maps and facilitate access to their rich information content.

关键词： reference patterns two-dimensional correlation spectra correlation coefficient maps perturbation domain decomposition trajectory peaks

来源：评论

学校读者我要写书评

暂无评论

FILTERING OF NETWORK ADDRESSES IN REAL-TIME BY SEQUENTIAL-DECODING

引用

IEE PROCEEDINGS-E COMPUTERS AND DIGITAL TECHNIQUES 1988年第1期135卷 55-59页

作者： WOLSTENHOLME, P CERN Geneva Switzerland

The Fano algorithm for sequential decoding permits the simultaneous comparison of a serial bit pattern, with a large number of reference patterns and is applicable to the identification of addresses in local area networks. Such a requirement arises frequently in the design of bridges (filtered, buffered, repeaters) between networks, operating at the data link layer so as to be transparent to various higher-level protocols. The paper explains how the algorithm functions and how it can be implemented in hardware, operating in real time as the address bits arrive. Several specific applications are discussed.

关键词： serial bit pattern bridges protocols Computer communications local area networks network addresses filtering Fano algorithm Information theory sequential decoding decoding Local area networks reference patterns

来源：评论

学校读者我要写书评

暂无评论

reference TEMPLATE ADAPTATION IN SPEAKER-INDEPENDENT ISOLATED WORD SPEECH RECOGNITION

引用

ELECTRONICS LETTERS 1987年第24期23卷 1304-1305页

作者： MCINNES, FR JACK, MA Centre for Speech Technology Research University of Edinburg Edinburgh UK

A technique which permits the adaptation of reference patterns (templates) in isolated word speech recognition systems is described. Experimental results for supervised and unsupervised adaptation with speaker-indepen... 详细信息

关键词： Speech recognition speaker-independent initial templates supervised adaptation template adaptation unsupervised adaptation Speech and audio signal processing speech recognition isolated word speech recognition reference patterns

来源：评论

学校读者我要写书评

暂无评论

A novel method for in-vivo evaluation of finger kinematics including definition of healthy motion patterns

引用

CLINICAL BIOMECHANICS 2016年 31卷 47-58页

作者： Coupier, Jerome Hamoudi, Samir Telese-Izzi, Sonia Feipel, Veronique Rooze, Marcel Jan, Serge Van Sint Univ Libre Bruxelles Fac Med LABO Brussels Belgium Univ Libre Bruxelles Fac Motor Sci Funct Anat Lab Brussels Belgium

Background: Despite recent progress in motion capture technology, such as stereophotogrammetry based on the tracking of markers set on the subject, it remains challenging to develop a complete protocol for in-vivo functional evaluation of the hand. The current practical problems regarding small anatomical segments, such as the fingers, are mainly due to the high concentration of markers in a relatively reduced volume. Methods: This paper proposes a novel procedure for hand functional analysis by analysing finger behaviour along the main displacement plane simultaneously with combined motions. The objective was two-fold. For one thing, a novel data collection protocol was implemented, which includes specific setting of the motion capture system and was to create a reference database of a healthy sample for further clinical ithe development of finger marker clusters. The second purpose of this studynvestigation. Twenty healthy volunteers took part in the study. Analytical motions (flexion/extension and abduction/adduction) of all five fingers were recorded. Findings: Results showed good correspondence with the literature. Specific kinematic behaviour of each analysed joint is reported. Statistically significant differences were found between the right and left sides of the subjects for the flexion/extension movement only, between the finger joints and between the fingers for all movements. No significant difference was found between genders. A validation protocol was performed, which proved the validity of the presented methodology. Interpretation: The protocol appears suitable for further use in motion analysis and for musculoskeletal modelling of the hand. It will also be considered for clinical application. (C) 2015 Elsevier Ltd. All rights reserved.

关键词： Finger kinematics Functional evaluation reference patterns Analytical motion

来源：评论

学校读者我要写书评

暂无评论

FOURIER OPTICAL SIGNAL PROCESSORS

引用

PROCEEDINGS OF THE IEEE 1989年第10期77卷 1511-1527页

作者： FLANNERY, DL HORNER, JL USAF ROME AIR DEV CTR SOLID STATE SCI DIRECTORATE BEDFORD MA 01731 USA

Progress in Fourier optical processing techniques is reviewed. Particular emphasis is placed on real-time pattern recognition, which recently has received increased interest as the result of new filter formulations that can be implemented with existing spatial light modulators. Architectures for coherent optical correlation are reviewed, as are the recently reported phase-only filters. Smart filters that attack the inherent distortion sensitivity of correlation, including those suitable for implementation with phase-only modulation, are reviewed. Correlation experiments implementing real-time variation of both input and reference patterns are also reviewed. The potential for successful near-term application of these techniques is examined

关键词： Fourier optical processing techniques Fourier transform optics coherent optical correlation filter formulations inherent distortion sensitivity optical correlation optical information processing optical modulation optical signal processors pattern recognition phase-only filters phase-only modulation real-time pattern recognition reference patterns spatial filters spatial light modulators

来源：评论

学校读者我要写书评

暂无评论

Underside information on PD-CAAS and DIAAS, two seemingly simple protein quality criteria

引用

CAHIERS DE NUTRITION ET DE DIETETIQUE 2021年第2期56卷 102-110页

作者： Tessier, Romain Calvez, Juliane Gaudichon, Claire Univ Paris Saclay AgroParisTech INRAE UMR PNCA 16 Rue Claude Bernard F-75005 Paris France

PD-CAAS and DIAAS are indexes that simply reflect protein quality based on protein digestibility and the amino acid composition of the protein relatively to a reference composition. The calculation of these scores raises critical points, among which the amino acid pattern of the reference protein depending on the age range. Moreover, the factor 6.25 used to convert nitrogen to protein to quantify the protein amount in the protein ingredient generally overestimates the protein content, resulting in an underestimation of quality scores. Additionally, using free or "in-chain'' molecular weight to estimate amino acids content affects significantly the score. Hence, it is important to clearly specify theses parameters in the studies. Lastly, the model used to determine digestibility (Human, pig, rat) as well as the methodology (ileal vs fecal, assessment of endogenous losses vs isotopic tracers) lead to variability. This must be carefully considered in the understanding of the PD-CAAS and DIAAS and the comparison between protein sources. (C) 2021 Societe francaise de nutrition. Published by Elsevier Masson SAS. All rights reserved.

关键词： Chemical score reference patterns Digestibility Amino acids

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：