Computer Sciences Commons^™

Aurora Working Group: Dsr Front End Lvcsr Evaluation — Baseline Recognition System Description, Naveen Parihar, Joseph Picone

Publications

In this document we describe the features of the baseline system to be used in the Distributed Speech Recognition (DSR) front end large vocabulary continuous speech recognition (LVCSR) evaluations being conducted by the Aurora Working Group of the European Telecommunications Standards Institute (ETSI). The objective of these evaluations is to determine the robustness of different front ends for use in client/server type telecommunications applications. As such, our experiments are designed to test the following focus conditions on the DARPA Wall Street Journal (WSJ0) corpus using a 5000-word closed-loop vocabulary and a bigram language model:

1. Additive Noise: six noise conditions …

The Need For Small Learning Rates On Large Problems, Tony R. Martinez, D. Randall Wilson

Faculty Publications

In gradient descent learning algorithms such as error backpropagation, the learning rate parameter can have a significant effect on generalization accuracy. In particular, decreasing the learning rate below that which yields the fastest convergence can significantly improve generalization accuracy, especially on large, complex problems. The learning rate also directly affects training speed, but not necessarily in the way that many people expect. Many neural network practitioners currently attempt to use the largest learning rate that still allows for convergence, in order to improve training speed. However, a learning rate that is too large can be as slow as a learning …

Optimal Artificial Neural Network Architecture Selection For Bagging, Timothy L. Andersen, Tony R. Martinez, Michael E. Rimer

Faculty Publications

This paper studies the performance of standard architecture selection strategies, such as cost/performance and CV based strategies, for voting methods such as bagging. It is shown that standard architecture selection strategies are not optimal for voting methods and tend to underestimate the complexity of the optimal network architecture, since they only examine the performance of the network on an individual basis and do not consider the correlation between responses from multiple networks.

Write Once, Move Anywhere: Toward Dynamic Interoperability Of Mobile Agent Systems, Arne Grimstrup, Robert Gray, David Kotz, Thomas Cowin, Greg Hill, Niranjan Suri, Daria Chacon, Martin Hofmann

Computer Science Technical Reports

Mobile agents are an increasingly popular paradigm, and in recent years there has been a proliferation of mobile-agent systems. These systems are, however, largely incompatible with each other. In particular, agents cannot migrate to a host that runs a different mobile-agent system. Prior approaches to interoperability have tried to force agents to use a common API, and so far none have succeeded. Our goal, summarized in the catch phrase ``Write Once, Move Anywhere,'' led to our efforts to develop mechanisms that support dynamic runtime interoperability of mobile-agent systems. This paper describes the Grid Mobile-Agent System, which allows agents to migrate …

Improving The Hopfield Network Through Beam Search, Tony R. Martinez, Xinchuan Zeng

Faculty Publications

In this paper we propose a beam search mechanism to improve the performance of the Hopfield network for solving optimization problems. The beam search readjusts the top M (M > 1) activated neurons to more similar activation levels in the early phase of relaxation, so that the network has the opportunity to explore more alternative, potentially better solutions. We evaluated this approach using a large number of simulations (20,000 for each parameter setting), based on 200 randomly generated city distributions of the 10-city traveling salesman problem. The results show that the beam search has the capability of significantly improving the network …

Improved Hopfield Networks By Training With Noisy Data, Fred Clift, Tony R. Martinez

Faculty Publications

A new approach to training a generalized Hopfield network is developed and evaluated in this work. Both the weight symmetricity constraint and the zero self-connection constraint are removed from standard Hopfield networks. Training is accomplished with Back-Propagation Through Time, using noisy versions of the memorized patterns. Training in this way is referred to as Noisy Associative Training (NAT). Performance of NAT is evaluated on both random and correlated data. NAT has been tested on several data sets, with a large number of training runs for each experiment. The data sets used include uniformly distributed random data and several data sets …

Lazy Training: Improving Backpropagation Learning Through Network Interaction, Timothy L. Andersen, Tony R. Martinez, Michael E. Rimer

Faculty Publications

Backpropagation, similar to most high-order learning algorithms, is prone to overfitting. We address this issue by introducing interactive training (IT), a logical extension to backpropagation training that employs interaction among multiple networks. This method is based on the theory that centralized control is more effective for learning in deep problem spaces in a multi-agent paradigm. IT methods allow networks to work together to form more complex systems while not restraining their individual ability to specialize. Lazy training, an implementation of IT that minimizes misclassification error, is presented. Lazy training discourages overfitting and is conducive to higher accuracy in multiclass problems …

Speed Training: Improving The Rate Of Backpropagation Learning Through Stochastic Sample Presentation, Timothy L. Andersen, Tony R. Martinez, Michael E. Rimer

Faculty Publications

Artificial neural networks provide an effective empirical predictive model for pattern classification. However, using complex neural networks to learn very large training sets is often problematic, imposing prohibitive time constraints on the training process. We present four practical methods for dramatically decreasing training time through dynamic stochastic sample presentation, a technique we call speed training. These methods are shown to be robust to retaining generalization accuracy over a diverse collection of real world data sets. In particular, the SET technique achieves a training speedup of 4278% on a large OCR database with no detectable loss in generalization.

Student Survey Of Information Technology, Wku Information Technology, Richard Kirchmeyer

Board of Regents Documents

Survey of 400 WKU students about information technology used in strategic operations planning for Information Technology. The survey attempted to determine the depth and breadth of student computer use and knowledge of a variety of hardware and software. The report was presented to the WKU Board of Regents at the August 17, 2001 meeting.

Precise Environmental Searches: Integrating Hierarchical Information Search With Envirodaemon, George Chang, Gunjan Samtani, Marcus Healey, Franz J. Kurfess, Jason Wang

Computer Science and Software Engineering

Information retrieval has evolved from searches of references, to abstracts, to documents. Search on the Web involves search engines that promise to parse full-text and other files: audio, video, and multimedia. With the indexable Web at 320 million pages and growing, difficulties with locating relevant information have become apparent. The most prevalent means for information retrieval relies on syntax-based methods: keywords or strings of characters are presented to a search engine, and it returns all the matches in the available documents. This method is satisfactory and easy to implement, but it has some inherent limitations that make it unsuitable for …

Semantic Operators And Fixed-Point Theory In Logic Programming, Anthony K. Seda, Pascal Hitzler

Computer Science and Engineering Faculty Publications

We consider rather general operators mapping valuations to (sets of) valuations in the context of the semantics of logic programming languages. This notion generalizes several of the standard operators encountered in this subject and is inspired by earlier work of M.C. Fitting. The fixed points of such operators play a fundamental role in logic programming semantics by providing standard models of logic programs and also in determining the computability properties of these standard models. We discuss some of our recent work employing topological ideas, in conjunction with order theory, to establish methods by which one can find the fixed points …

Neural Network Approach To Causal Reasoning With Penalty Logic, Ghada Moussa Abdel Ghany Bahig

Archived Theses and Dissertations

No abstract provided.

Moving Icons, Detection And Distraction, Lyn Bartram, Colin Ware, Tom Calvert

Center for Coastal and Ocean Mapping

Simple motion has great potential for visually encoding information but there are as yet few experimentally validated guidelines for its use. Two studies were carried out to look at how efficiently simple motion cues were detected and how distracting they were in different task contexts. The results show that motion outperforms static representations and identify certain types of motions which are more distracting and irritating than others.

Development Of Atlas Based Simulation Capability For Automated Testing, Rami Hanbali

Electrical & Computer Engineering Theses & Dissertations

Computer software is emerging as a powerful tool for controlling a large number of instruments and for the testing of these instruments. The main aim of this work is to provide a software program capable of controlling a large number of engineering instruments at the touch of a button. In addition, the software is to have the capability of connecting the instruments with the desired Unit Under Test. There is great need for such software driven testing in industry, especially for large complex systems. Advantages of such an approach include: (a) automated testing in a very quick and efficient manner; …

Minimum Mean Square Error Spectral Peak Envelope Estimation For Automatic Vowel Classification, Jaishree Venugopal

Electrical & Computer Engineering Theses & Dissertations

Spectral feature computations continue to be a very difficult problem for accurate machine recognition of speech. In this work, which focuses on vowels, a new spectral peak envelope method for vowel classification is developed, based on a missing frequency components model of speech recognition. According to the missing frequency components model, vowel recognition depends only on the spectral (harmonic) peaks. Smoothing and interpolation of the spectra, performed in the standard cepstral analysis method commonly used in automatic speech recognition, actually loses valuable information and results in reduced recognition accuracy. The new method for feature extraction presented in this thesis is …

Web Spoofing 2001, Yougu Yuan, Eileen Zishuang Ye, Sean Smith Dartmouth College

Computer Science Technical Reports

The Web is currently the pre-eminent medium for electronic service delivery to remote users. As a consequence, authentication of servers is more important than ever. Even sophisticated users base their decision whether or not to trust a site on browser cues---such as location bar information, SSL icons, SSL warnings, certificate information, response time, etc. In their seminal work on web spoofing, Felten et al showed how a malicious server could forge some of these cues---but using approaches that are no longer reproducible. However, subsequent evolution of Web tools has not only patched security holes---it has also added new technology to …

Securing Web Servers Against Insider Attack, Shan Jiang, Sean Smith, Kazuhiro Minami Dartmouth College

Computer Science Technical Reports

Too often, ``security of Web transactions'' reduces to ``encryption of the channel''---and neglects to address what happens at the server on the other end. This oversight forces clients to trust the good intentions and competence of the server operator---but gives clients no basis for that trust. Furthermore, despite academic and industrial research in secure coprocessing, many in the computer science community still regard ``secure hardware'' as a synonym for ``cryptographic accelerator.' This oversight neglects the real potential of COTS secure coprocessing technology to establish trusted islands of computation in hostile environments---such as at web servers with risk of insider attack. …

An Information Theoretic Methodology For Prestructuring Neural Networks, Bjorn Chambless, George G. Lendaris, Martin Zwick

Complex Systems Faculty Publications and Presentations

Absence of a priori knowledge about a problem domain typically forces use of overly complex neural network structures. An information-theoretic method based on calculating information transmission is applied to training data to obtain a priori knowledge that is useful for prestructuring (reducing complexity) of neural networks. The method is applied to a continuous system, and it is shown that such prestructuring reduces training time, and enhances generalization capability.

Cost Optimal Record/Entity Matching, V. S. Verykios, Ahmed K. Elmagarmid, G. V. Moustakides

Department of Computer Science Technical Reports

No abstract provided.

Morphological And Physiological Effects Of Mechanical Trauma To The Spinal Cord, Steven Teoh, Yinlong Sun

Department of Computer Science Technical Reports

No abstract provided.

Record Matching: Past, Present And Future, M. Cochinwala, S. Dalal, Ahmed K. Elmagarmid, V. S. Verykios

Department of Computer Science Technical Reports

No abstract provided.

Broadcasting Indexed Multidimensional Data, Susanne E. Hambrusch, Chuan-Ming Liu, Walid G. Aref, Sunil Prabhakar

Department of Computer Science Technical Reports

No abstract provided.

Query Indexing And Velocity Constrained Indexing: Scalable Techniques For Continuous Queries On Moving Objects, Sunil Prabhakar, Y. Xia, D. Kalashnikov, Walid G. Aref, Susanne E. Hambrusch

Department of Computer Science Technical Reports

No abstract provided.

On The Utility Of Entanglement In Quantum Neural Computing, Dan A. Ventura

Faculty Publications

Efforts in combining quantum and neural computation are briefly discussed and the concept of entanglement as it applies to this subject is addressed. Entanglement is perhaps the least understood aspect of quantum systems used for computation, yet it is apparently most responsible for their computational power. This paper argues for the importance of understanding and utilizing entanglement in quantum neural computation.

An Evaluation Of Shared Multicast Trees With Multiple Active Cores, Daniel Zappala, Aaron Fabbri

Faculty Publications

Core-based multicast trees use less router state, but have significant drawbacks when compared to shortest-path trees, namely higher delay and poor fault tolerance. We evaluate the feasibility of using multiple independent cores within a shared multicast tree. We consider several basic designs and discuss how using multiple cores improves fault tolerance without sacrificing router state. We examine the performance of multiple-core trees with respect to single-core trees and find that adding cores significantly lowers delay without increasing cost. Moreover, it takes only a small number of cores, placed with a k-center approximation, for a multiple-core tree to have lower delay …

Using Mobile Agents For Analyzing Intrusion In Computer Networks, Jay Aslam, Marco Cremonini, David Kotz, Daniela Rus

Dartmouth Scholarship

Today hackers disguise their attacks by launching them form a set of compromised hosts distributed across the Internet. It is very difficult to defend against these attacks or to track down their origin. Commercially available intrusion detection systems can signal the occurrence of limited known types of attacks. New types of attacks are launched regularly but these tools are not effective in detecting them. Human experts are still the key tool for identifying, tracking, and disabling new attacks. Often this involves experts from many organizations working together to share their observations, hypothesis, and attack signatures. Unfortunately, today these experts have …

Recent Advances In Content-Based Video Analysis, Chong-Wah Ngo, Ting-Chuen Pong, Hong-Jiang Zhang

Research Collection School Of Computing and Information Systems

In this paper, we present major issues in video parsing, abstraction, retrieval and semantic analysis. We discuss the success, the difficulties and the expectations in these areas. In addition, we identify important opened problems that can lead to more sophisticated ways of video content analysis. For video parsing, we discuss topics in video partitioning, motion characterization and object segmentation. The success in video parsing, in general, will have a great impact on video representation and retrieval. We present three levels of abstracting video content by scene, keyframe and key object representations. These representation schemes in overall serve as a good …

Secure And Private Distribution Of Online Video And Some Related Cryptographic Issues, Feng Bao, Robert H. Deng, Peirong Bao, Yan Guo, Hongjun Wu

Research Collection School Of Computing and Information Systems

With the rapid growth of broadband infrastructure, it is thought that the bottleneck for video-on-demand service through Internet is being cleared. However, digital video content protection and consumers privacy protection emerge as new major obstacles. In this paper we propose an online video distribution system with strong content security and privacy protection. We mainly focus on the study of security and privacy problems related to the system. Besides presenting the new system, we intensively discuss some relevant cryptographic issues, such as content protection, private information retrieval, super-speed encryption/decryption for video, and PKC with fast decryption etc. The paper can be …

Mobile Commerce: Promises, Challenges And Research Agenda, Keng Siau, Ee Peng Lim

Research Collection School Of Computing and Information Systems

Advances in wireless technology increase the number of mobile device users and give pace to the rapid development of e-commerce using these devices. The new type of e-commerce, conducting transactions via mobile terminals, is called mobile commerce. Due to its inherent characteristics such as ubiquity, personalization, flexibility, and dissemination, mobile commerce promises businesses unprecedented market potential, great productivity, and high profitability. This paper presents an overview of mobile commerce development by examining the enabling technologies, the impact of mobile commerce on the business world, and the implications to mobile commerce providers. The paper also provides an agenda for future research …

Application-Centric Analysis Of Ip-Based Mobility Management Techniques, Archan Misra, Subir Das, Prathima Agrawal

Research Collection School Of Computing and Information Systems

This paper considers three applications—VoIP, mobile Web access and mobile server-based data transfers—and evaluates the applicability of various IP-based mobility management mechanisms. We first survey the features and characteristics of various IP mobility protocols, such as MIPv4, MIPv6, MIP-RO, SIP, CIP, HAWAII, MIP-RR and IDMP, and then evaluate their utility on an application-specific basis. The diversity in the mobility-related requirements ensures that no single mobility solution is universally applicable. We recommend a hierarchical mobility architecture. The framework uses our Dynamic Mobility Agent (DMA) architecture for managing intra-domain mobility and multiple application-based binding protocols for supporting inter-domain mobility. Thus, we recommend …