检索结果-内蒙古大学图书馆

International Conference on Spoken Language, ICSLP

作者： M.J. Adamson R.I. Damper Image Speech and Intelligent Systems (ISIS) Research Group Department of Electronics and Computer Science University of Southampton Southampton UK

Previous attempts to derive connectionist models for text-to-phoneme conversion-such as NETtalk and NETspeak-have generally used pre-aligned training data and purely feedforward networks, both of which represent simplifications of the problem. In this work, we explore the potential of recurrent networks to perform the conversion task when trained on non-aligned data. Initially, our use of a single recurrent network produced disappointing results. This led to the definition of a two-phase model in which the hidden-unit representation of an auto-associative network was fed forward to a recurrent network. Although this model currently does not perform as well as NETspeak it is solving a harder problem. Also, we propose several possible avenues for improvement.

关键词： Training data Shock absorbers Damping Speech Intelligent systems Intelligent networks Intersymbol interference computer science Image converters Multilayer perceptrons

来源：评论

学校读者我要写书评

暂无评论

Identification of boundaries in MRI medical images using artificial neural networks

Identification of boundaries in MRI medical images using art...

引用

IEE Colloquium on Artificial Intelligence Methods for Biomedical Data Processing

作者： I. Middleton R.I. Damper Dept. of Electron. & Comput. Sci. Southampton Univ. UK Image Speech and Intelligent Systems Research Group Department of Electronics and Computer Science University of Southampton Southampton UK

In the area of medical imaging, fully-automatic and robust segmentation techniques would have an enormous beneficial impact on clinical practice and research, by decreasing dramatically the manual effort which must otherwise be devoted to this task. Deployment of conventional image processing techniques has not so far led to a fully-automatic solution, although semi-automatic systems do exist. Since no known, robust segmentation algorithm exists, the ability of neural networks to discover regularities and features in complex data is appealing. Indeed, many preliminary attempts at neural segmentation have been described, although none yet achieves the necessary level of performance for routine application. Southampton General Hospital have a requirement to obtain lung-boundary data within an asthma research project. In connection with this requirement, we have previously reported on work in which multilayer perceptrons (MLPs) are trained using backpropagation to segment the region of the lungs in magnetic resonance images of the thorax. This is achieved by training the network to classify voxels as either boundary (voxels on the boundary between lung interior and surrounding tissue) or non-boundary. In this paper, we present the latest results using this technique. We also show how the generalisation performance of the MLP can be improved using a variety of techniques, including weight pruning algorithms.

关键词： Biomedical magnetic resonance imaging Respiratory system Biomedical image processing Image classification Multilayer perceptrons Backpropagation Artificial intelligence Image edge analysis

来源：评论

学校读者我要写书评

暂无评论

Roles of neural and evolutionary computing in intelligent software systems

British Telecom technology journal

引用

British Telecom technology journal 1996年第4期14卷 46-54页

作者： Tsui, K.C. Azvine, B. Plumbley, M. Hong Kong Polytechnic University of Essex Kings College University of London BT Laboratories Manchester University Dynamics and Control Research Group Engineering Department Manchester University Contr. and Instrum. Research Group Staffordshire University Intelligent Systems Research Group Churchill College Cambridge United Kingdom Thorn-EMI Ctrl. Res. Laboratories Cambridge Univ. Eng. Department Centre for Neural Networks King's College London Department of Computer Science Electron. and Electrical Engineering King's College London United Kingdom European Neural Network Society

This paper describes two computing paradigms known as neural computing and evolutionary computing, and their potential contribution to building intelligent software systems. The paper begins by giving a brief introduction to the origins of each paradigm. Then two sections introduce the basic principles, and identify the role of each paradigm in intelligent system design. Each section ends with a number of applications that have been or are being investigated. These include connection admission control, modem communication, adaptive model-based control, face and handwriting recognition, frequency assignment, help-desk scheduling, financial time-series prediction, face recognition and evolving agent behavior. A section introduces the idea of using communications theory to design neural networks and the paper concludes with the authors' views on the future of neural and evolutionary computing for intelligent software systems.

关键词： Artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

A GLOBAL GRADIENT-NOISE COVARIANCE EXPRESSION FOR STATIONARY REAL GAUSSIAN INPUTS

引用

IEEE TRANSACTIONS ON NEURAL networks 1995年第6期6卷 1549-1551页

作者： AN, PE BROWN, M HARRIS, CJ Advanced Systems Research Group Department of Electronics and Computer Science University of Southampton Southampton UK

Supervised parameter adaptation in many artificial neural networks is largely based on an instantaneous version of gradient descent called the least-mean-square (LMS) algorithm, As the gradient is estimated using single samples of the input ensemble, its convergence properties generally deviate significantly from that of the true gradient descent because of the noise in the gradient estimate, It is thus important to study the gradient-noise characteristics so that the convergence of the LMS algorithm can be analyzed from a new perspective, This paper considers only neural models which are Linear with respect to their adaptable parameters and has two major contributions, First, it derives an expression for the gradient-noise covariance under the assumption that the input samples are real, stationary, Gaussian distributed;but can be partially correlated, This expression relates the gradient correlation and input correlation matrices to the gradient-noise covariance and explains why the gradient noise generally correlates maximally with the steepest principal axis and minimally with the one of the smallest curvature, regardless of the magnitude of the weight error, Second, a recursive expression for the weight-error correlation matrix is derived ina straightforward manner using the gradient-noise covariance, and comparisons are drawn with the complex LMS algorithm.

关键词： Least squares approximation Convergence Covariance matrix Artificial neural networks Iterative algorithms Parameter estimation Algorithm design and analysis Gaussian noise Large-scale systems Training data

来源：评论

学校读者我要写书评

暂无评论

PROVING PROPERTIES OF REAL-TIME SEMAPHORES

引用

science OF computer PROGRAMMING 1995年第2期24卷 159-181页

作者： SCHOLEFIELD, D Formal Systems Research Group Department of Computer Science University of York Heslington York YO1 5DD UK

Much work has been undertaken on investigating the use of semaphore primitives in concurrent programming languages. It has been shown that semaphores are adequate for expressing many forms of concurrency control, including the enforcement of communication protocols, and mutual exclusion protocols on shared resources. In this paper we present a formal language for real-time distributed programs which includes a semaphore primitive. This primitive is used to lock and unlock resources which are directly associated with either processors or communication channels. The semaphores are real-time, i.e. the programmer can express timing constraints about when the semaphores should lock and unlock. It is demonstrated that, using these semaphores, a number of apparently disjoint issues in real-time distributed systems theory can be unified within a single notion of resource restriction. In particular it is shown that different models of communication, control of shared access to resources (mutual exclusion), and process to processor mapping (physical placement), can all be expressed and reasoned about in a unified manner.

关键词： FORMAL VERIFICATION REAL-TIME systems SEMAPHORES LIMITED RESOURCES COMMUNICATION CONCURRENCY SCHEDULING

来源：评论

学校读者我要写书评

暂无评论

Object-oriented programming with protected types in Ada 95 3

Object-oriented programming with protected types in Ada 95

引用

3rd Workshop on Parallel and distributed Real-Time systems, WPDRTS 1995

作者： Wellings, A.J. Mitchell, S. Burns, A. Real-time Systems Research Group Department of Computer Science University of York York United Kingdom

ISBN: (纸本)0818670991

Integrating concurrent and object-oriented language facilities is currently an active research area. There are a few experimental languages which attempt to combine various models of concurrency within an OOP framework. Most of these suffer from the so called inheritance anomaly where a concurrent object's synchronisation code needs to be modified if the object is extended. Ada 95 has avoided some of these problems by integrating protected types into the model of type extensibility. However this has been at the expense of some loss of generality. This paper considers three paradigms of combining protected types with the type extension. The strengths and weaknesses of each approach is discussed. We conclude that none of the approaches are entirely satisfactory and they either suffer from the inheritance anomaly or preclude further extensions. © 1995 IEEE.

关键词： Object oriented programming

来源：评论

学校读者我要写书评

暂无评论

Guaranteeing Synchronous Message Sets in FDDI networks

引用

IFAC Proceedings Volumes 1995年第22期28卷 105-110页

作者： Sijing Zhang Alan Burns Real-Time Systems Research Group Department of Computer Science University of York UK

This paper addresses the issue related to the deadline guarantees of synchronous messages in FDDI networks. The exact value of X i (the minimum amount of time for each node i on the network to transmit its synchronous messages during the message’s (relative) deadline D i ), which is better than the previously published, is derived. Based on this result, the schedulability of synchronous message sets is generally analysed. This analysis differs from previous ones by taking into account synchronous message sets with the minimum message deadline ( D min ) less than twice the Target Token Rotation Time (TTRT).

关键词： Real-time communication token-ring protocol bandwidth allocation local area networks communication protocols

来源：评论

学校读者我要写书评

暂无评论

Recognition and anticipation of hand motions using a recurrent neural network

Recognition and anticipation of hand motions using a recurre...

引用

International Conference on Neural networks

作者： P. Vamplew A. Adams Artificial Neural Networks Research Group Department of Computer Science University of Tasmania Australia

来源：评论

学校读者我要写书评

暂无评论

Secure pay-per-view testbed

Secure pay-per-view testbed

引用

IEEE International Conference on Multimedia Computing and systems (ICMCS)

作者： D. Cunningham D. O'Mahony Networks and Telecommunications Research Group Department of Computer Science Trinity College Dublin Dublin Ireland

Trials of video on demand and pay-per-view systems are underway in many countries, and as a result, many cable and telecommunications companies are having to upgrade or replace their distribution networks. Video streams must be secure in order to prevent unauthorized viewing of the programs being transmitted, yet most existing security systems do not make use of the potential for bidirectional signalling in new and upgraded networks. It would prove more useful to many companies if they could use one system to test a number of different distribution networks to see if they save an acceptable quality of service. In this manner, a company could decide if existing networks could be used without substantial changes, or whether expensive upgrades or indeed replacements were justified.

关键词： Testing Video compression Video on demand Bidirectional control Watches Cable TV Streaming media Smart cards Optical films Transform coding

来源：评论

学校读者我要写书评

暂无评论

Evolving recurrent neural networks with non-binary encoding

Evolving recurrent neural networks with non-binary encoding

引用

IEEE International Conference on Evolutionary Computation

作者： M. Mandischer Systems Analysis Research Group Department of Computer Science University of Dortmund Dortmund Germany

This paper presents an evolutionary approach for the design of feedforward and recurrent neural networks. We show that evolutionary algorithms can be used for the construction of networks for real-world tasks. Therefore, a data structure based genotypic network representation, as well as genetic operators, are introduced. Results from the classification, function approximation and time-series domains are presented.

关键词： Recurrent neural networks Encoding Neural networks Feedforward systems Evolutionary computation Genetic algorithms Network topology computer science Electronic mail Function approximation

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：