پایان نامه های شبکه عصبی مصنوعی(7)

Dawn glimmers for day of the man-made brain
Froelich, Warren . The San Diego Union ; San Diego, Calif. [San Diego, Calif]06 July 1986: A-1.
ProQuest document link




ABSTRACT (ABSTRACT)
What [Robert Hecht-Nielsen] and others are talking about are networks that are capable of learning, whose architecture -- or operation design - - consists of vast interconnections similar to those found in the brain. In many ways, these networks are similar to what people think of as artificial intelligence -- or a system of machines that think.
At UCSD, researchers led by Dr. David Rumelhart, a psychologist, are using neural networks to study how humans process language, teaching it grammar such as changing verbs from present to past tense. Neural networks also could have several other practical applications, ranging from industrial inspection to the diagnosis of medical disorders.
Both [David Zipser] and Hecht-Nielsen emphasized that the present generation of neural networks may or may not do what the brain does. Critics have challenged the work in the past, saying the brain is too complex to emulate.

LINKS
DETAILS

Company / organization: Name: TRW Inc; Ticker: TRW; NAICS: 334419, 336399, 333911, 561450, 336414; SIC: 3675; DUNS: 00-417-9453

Publication title: The San Diego Union; San Diego, Calif.

Pages: A-1

Number of pages: 0

Publication year: 1986

Publication date: Jul 6, 1986

Dateline: RANCHO CARMEL

Section: NEWS

Publisher: The San Diego Union-Tribune, LLC.

Place of publication: San Diego, Calif.

Country of publication: United States

Publication subject: General Interest Periodicals--United States

Source type: Newspapers

Language of publication: English

Document type: NEWSPAPER

ProQuest document ID: 422526896

Document URL: https://search.proquest.com/docview/422526896?accountid=8243

Copyright: Copyright Union-Tribune Publishing Co. Jul 6, 1986

Last updated: 2010-08-20

Database: Global Newsstream



'Wise' Computer To Have Impact In Workplace
Anderson, Julie . Omaha World - Herald ; Omaha, Neb. [Omaha, Neb]12 Mar 1989: 1g.
ProQuest document link




ABSTRACT (ABSTRACT)
Alvin Surkan, a University of Nebraska-Lincoln computer science professor, is working to transform that dream from science fiction to science fact.
Surkan said computers one day could change the way employees are hired, by better matching "the right people to the right positions" and watching for those people who could prove to be security risks if placed in sensitive jobs.
Surkan said that using networks in hiring could benefit the hiring firm and ensure greater job satisfaction among workers. Networks also could be used to identify possible security risks in banking and analyze workers in hospitals and institutions where lives are placed in the hands of medical personnel. Entering Industry

LINKS
DETAILS

Publication title: Omaha World - Herald; Omaha, Neb.

Pages: 1g

Number of pages: 0

Publication year: 1989

Publication date: Mar 12, 1989

Section: Working

Publisher: Omaha World-Herald Company

Place of publication: Omaha, Neb.

Country of publication: United States

Publication subject: General Interest Periodicals--United States

Source type: Newspapers

Language of publication: English

Document type: NEWSPAPER

ProQuest document ID: 400631952

Document URL: https://search.proquest.com/docview/400631952?accountid=8243

Copyright: (Copyright 1989 Omaha World-Herald Company)

Last updated: 2011-10-19

Database: Global Newsstream



Artificial neural networks for pattern recognition in patient pain drawings
Mann, Noah Horace, III . University of Miami, ProQuest Dissertations Publishing, 1990. 9104436.
ProQuest document link




ABSTRACT
The purpose of this research is to study the pattern recognition capabilities of artificial neural networks on quantitative spatial-anatomic data obtained from the pain drawings of low back pain patients.
Experiments are performed with a self-correcting artificial intelligence program to determine if they can be used to categorize patient pain drawings into clinically relevant classes. The study is designed to demonstrate the capabilities of fundamental artificial neural network theory for (A) recognizing patterns in empirical patient data, and (B) using knowledge communicated by experts on a pragmatic, ill-defined medical application. Also, the analysis increases the current body of knowledge regarding the quantitative spatial-anatomic characteristics of pain drawing measurements.
Our results indicate that artificial neural networks can achieve classification accuracy comparable to human experts and traditional discriminant analysis without the overhead associated with the latter.
Artificial neural systems yield useful, often neglected, diagnostic information through their topology, and may assist in producing more reliable and more comprehensive outcomes from other diagnostic systems. The most accurate ANN performed with 48% overall accuracy compared to 51% for human experts and 47.4% for traditional discriminant analysis.

LINKS
DETAILS

Subject: Biomedical research; Electrical engineering; Computer science

Classification: 0541: Biomedical research; 0544: Electrical engineering; 0984: Computer science

Identifier / keyword: Applied sciences neural networks

Number of pages: 270

Publication year: 1990

Degree date: 1990

School code: 0125

Source: DAI-B 51/09, Dissertation Abstracts International

Place of publication: Ann Arbor

Country of publication: United States

Advisor: Brown, Mark D. Hertz, David B.

University/institution: University of Miami

University location: United States -- Florida

Degree: Ph.D.

Source type: Dissertations &Theses

Language: English

Document type: Dissertation/Thesis

Dissertation/thesis number: 9104436

ProQuest document ID: 303825279

Document URL: https://search.proquest.com/docview/303825279?accountid=8243

Copyright: Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.

Database: ProQuest Dissertations &Theses Global



Analysis and applications of general classes of dynamic neural networks
Farotimi, Oluseyi Oladele . Stanford University, ProQuest Dissertations Publishing, 1990. 9102260.
ProQuest document link




ABSTRACT
Research interest in neural networks has grown over the past few years in the hope that they may offer more efficient alternatives to conventional algorithms. Generally speaking, along the path from research to development two main issues arise, namely (i) qualitative behavior of the systems, and (ii) training rules. Qualitative analysis of first order networks has been carried out by Cohen and Grossberg, among others. Widrow, Rumelhart, Hopfield, and others have proposed various training rules for different network structures.
In this thesis results pertaining to training as well as to qualitative analysis of neural networks are presented. In some cases they represent generalizations of existing results, and in other cases they introduce entirely novel concepts.
First, a new technique for training neural networks based on optimal control theory is presented. This method is different from many existing rules in that it places very few constraints on the order or architecture of the network. The method yields an optimal weight matrix that is a function of time.
The optimal control technique is applied to train the weights in an associative memory. For this problem, a common weight rule is the outer product rule, introduced by Hopfield. By considering special cases of the performance index, optimal rules for the problem are derived, and encouraging simulation results are presented.
Still addressing the issue of neural network training, the optimal control technique above is applied to determine the weights in a Probabilistic Cellular Automaton (PCA) for pattern recognition. Two ways of determining the weights in this structure are examined, and simulation results are presented for some simple examples.
Finally, a qualitative analysis of a class of arbitrary order dynamic neural networks is presented. Such networks at steady state can give rise to polynomial threshold functions (Bruck 1989). Other applications for such networks include higher order associative memories and nonlinear programming. All these applications place certain constraints on the nature of the equilibrium points of the neural network. The analysis characterizes these equilibrium points.

LINKS
DETAILS

Subject: Electrical engineering; Computer science; Neurology

Classification: 0544: Electrical engineering; 0984: Computer science; 0317: Neurology

Identifier / keyword: Applied sciences Biological sciences pattern recognition

Number of pages: 161

Publication year: 1990

Degree date: 1990

School code: 0212

Source: DAI-B 51/08, Dissertation Abstracts International

Place of publication: Ann Arbor

Country of publication: United States

Advisor: Kailath, Thomas

University/institution: Stanford University

University location: United States -- California

Degree: Ph.D.

Source type: Dissertations &Theses

Language: English

Document type: Dissertation/Thesis

Dissertation/thesis number: 9102260

ProQuest document ID: 303871572

Document URL: https://search.proquest.com/docview/303871572?accountid=8243

Copyright: Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.

Database: ProQuest Dissertations &Theses Global



REPORT ON THE OFFICE AND COMPUTERS Playing mind games Neural network imitates brain
Zeidenberg, Jerry . The Globe and Mail ; Toronto, Ont. [Toronto, Ont]23 Oct 1990: C.1.
ProQuest document link




ABSTRACT (ABSTRACT)
"There's half a billion pap smears done each year, and all are checked out by hand," Mr. [Geoffery Hinton] said. "By letting the neural network do it, you save a lot of people from a lot of boring, routine work."


LINKS
DETAILS

Publication title: The Globe and Mail; Toronto, Ont.

Pages: C.1

Number of pages: 0

Publication year: 1990

Publication date: Oct 23, 1990

Publisher: The Globe and Mail

Place of publication: Toronto, Ont.

Country of publication: Canada

Publication subject: General Interest Periodicals--Canada

ISSN: 03190714

Source type: Newspapers

Language of publication: English

Document type: NEWSPAPER

ProQuest document ID: 385909988

Document URL: https://search.proquest.com/docview/385909988?accountid=8243

Copyright: All material copyright Bell Globemedia Publishing Inc. or its licensors. All rights reserved.

Last updated: 2017-11-08

Database: Global Newsstream



A "neural-RISC" processor and parallel architecture for neural networks
Pacheco, Marco Aurelio Cavalcanti . University of London, University College London (United Kingdom), ProQuest Dissertations Publishing, 1991. 10608847.
ProQuest document link




ABSTRACT (ENGLISH)
This thesis investigates a RISC microprocessor and a parallel architecture designed to optimise the computation of neural network models. The "Neural-RISC" is a primitive transputer-like microprocessor for building a parallel MIMD (multiple instruction, multiple data) general-purpose neurocomputer. The thesis covers four major parts: the design of the Neural-RISC system architecture, the design of the Neural-RISC node architecture, the architecture simulation studies, and the VLSI implementation of a microchip prototype. The Neural-RISC system architecture consists of linear arrays of microprocessors connected in rings. Rings end up in an interconnecting module forming a cluster. Clusters of rings are arranged in different point-to-point topologies and are controlled by a host computer. The interconnect module in each cluster acts as a communications server supporting inter-ring and inter-cluster message routing. The host, which consists of a workstation, supports network initialisation, programming and monitoring. During operation, messages in the form of packets can address: a node, a distinct group of nodes (cf. a neural network layer or cluster), all nodes (cf. broadcast), or the host. The neurocomputer nodes are configurated by downloading simple programs into each microprocessor. The Neural-RISC node architecture comprises a 16-bit reduced instruction-set processor, a communication unit, and local memory-all integrated into the same silicon die. The processor employs 16 instructions: 11 execute in one cycle; 4 in two cycles, and the multiply instruction executes in 16 cycles. One expanding opcode branches into a set of single-cycle, memory-mapped instructions. The communication unit provides four (unidirectional) point-to-point 16-bit links and a simple protocol for routing packets. Local memory contains: a RAM memory for instructions and data; two variable length FIFO buffers (as part of the working memory) to support the communication links; and a bootstrapping ROM. The architecture simulation studies involved the development of a software simulator and a simulation environment which entirely covered all steps in the process of programming and executing neural network models. The architecture simulator was implemented in C to aid in the design choices and to assess the proposed system. A clock-driven, register- level simulator realises each component of the Neural-RISC (system and node) architecture as configurable modules. Using the simulation environment, neural network models, written in the neural network implementation language NIL, were compiled, mapped and executed, to evaluate the system's performance, the network addressing scheme and the processor's instruction set. A VLSI prototype chip was implemented to demonstrate the system and node architecture. Using the standard 2μ CMOS technology, the chip integrates an array of two Neural-RISC microprocessors. Statistical analysis based on its results, provided an assessment of the chip's packing density and performance, and the hardware requirements of complete neurocomputer systems, for Neural-RISC chips implemented with modern CMOS technologies. Chip implementation involved the design of a customised datapath cell library, and independent PLA driven controllers for the processor and the communication units.

LINKS
DETAILS

Subject: Artificial intelligence; Computer science

Classification: 0800: Artificial intelligence; 0984: Computer science

Identifier / keyword: (UMI)AAI10608847 Applied sciences Parallel architecture

Number of pages: 188

Publication year: 1991

Degree date: 1991

School code: 6022

Source: DAI-C 75/03, Dissertation Abstracts International

Place of publication: Ann Arbor

Country of publication: United States

ISBN: 9781369799514

University/institution: University of London, University College London (United Kingdom)

Department: Department of Computer Science

University location: England

Degree: Ph.D.

Source type: Dissertations &Theses

Language: English

Document type: Dissertation/Thesis

Dissertation/thesis number: 10608847

ProQuest document ID: 1914307936

Document URL: https://search.proquest.com/docview/1914307936?accountid=8243

Copyright: Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.

Database: ProQuest Dissertations &Theses Global



Generalization in neural networks: Experiments in speech recognition
Richards, Elizabeth Lake . University of Colorado at Boulder, ProQuest Dissertations Publishing, 1991. 9206643.
ProQuest document link




ABSTRACT
This research is an investigation of the problem of generalization in neural networks: how do the task which the network must learn, the architecture of the network, the training of the network, and the data representations used in that training, both individually and collectively, affect the ability of a network to learn the training data and to generalize well to novel data.
A psychological model of speech perception, Liberman and Mattingly's Motor Theory, provides the theoretical foundation for the tasks and architectures specified for the networks used in the research. Linguistic theories of vowel perception guided the preparation of speech data representations used in training the networks. Vowel data was collected across varying contexts and speakers to provide a broad test of the networks' ability to generalize to highly variable data.
Results of the research show that networks having different task requirements but trained with the same number and type of data representations form a family of networks which exhibit similar generalization across a broad range of hidden units. Contradicting commonly accepted guidelines, networks trained with larger data representations exhibit better generalization than networks trained with smaller representations, even though the larger networks have a significantly greater capacity. In addition, networks having the same training performance can exhibit different levels of generalization; researchers interested in generalization must track generalization directly. Finally, given an appropriate architecture, training algorithm, and sufficient training data, the data representation itself is the primary determiner of a network's ability to generalize well to new data.

LINKS
DETAILS

Subject: Computer science; Artificial intelligence

Classification: 0984: Computer science; 0800: Artificial intelligence

Identifier / keyword: Applied sciences

Number of pages: 183

Publication year: 1991

Degree date: 1991

School code: 0051

Source: DAI-B 52/09, Dissertation Abstracts International

Place of publication: Ann Arbor

Country of publication: United States

Advisor: Bradshaw, Gary L.

University/institution: University of Colorado at Boulder

University location: United States -- Colorado

Degree: Ph.D.

Source type: Dissertations &Theses

Language: English

Document type: Dissertation/Thesis

Dissertation/thesis number: 9206643

ProQuest document ID: 303915881

Document URL: https://search.proquest.com/docview/303915881?accountid=8243

Copyright: Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.

Database: ProQuest Dissertations &Theses Global



Integrating neural networks with influence diagrams for multiple sensor diagnostic systems
Tseng, Ming-Lei . University of California, Berkeley, ProQuest Dissertations Publishing, 1991. 9228887.
ProQuest document link




ABSTRACT
The objective of this research is to develop an adaptive architecture for the fusion of sensory information for diagnostic reasoning. The system architecture envisioned uses influence diagrams to provide a symbolic representation of the manufacturing or process system model. The independence encoded in influence diagrams is utilized in defining the structure of the neural networks to reduce the system complexity. Neural networks are then used as a learning mechanism for estimating and updating conditional probabilities from a set of training data.
To this end, we have designed two types of probabilistic neural networks: probabilistic Hamming nets (PHNs) for discrete variables and self-organizing probabilistic neural networks (SOPNNs) for continuous variables. A PHN is basically a modified Hamming net which implements pattern matching in the first layer and estimates conditional probabilities in the second layer. This model uses a Bayesian learning rule to update the weights of networks. A SOPNN is a neural network with a fixed-size self-organizing net for clustering and a modified Specht's model for estimating the probability density functions (PDFs). The idea is to represent similar training patterns in the same cluster by its centroid and then use these centroids for PDF estimation.
Several sets of simulated and real sensor data were employed to test the effectiveness of the proposed models. The results were analyzed and compared with those obtained by the conjunctoid, Specht's model and the polynomial Adaline (Padaline) neural models as well as Bayesian learning rules. The effectiveness of the integrated network for reducing the system complexity both in representation and in inference is also demonstrated. Based on the results of this study, our proposed architecture has the following advantages: (1) The structure of the neural networks is well defined and less complex. (2) Our neural models (PHNs and SOPNNs) are of fixed-size. (3) Both models are very general, not limited to pre-specified parametric or unimodal distributions. (4) The SOPNN is self-organized. (5) The learning is one-pass and incremental. (6) Training and performance are very fast. (7) Partial data are better utilized. (8) Subjective information can be easily incorporated. (9) The integrated network can be flexibly reconfigured to respond to on-line requests. Since our models do not employ the time-consuming backpropagation and simulated annealing algorithms, this integrated network appears more promising for real-time applications.

LINKS
DETAILS

Subject: Mechanical engineering; Computer science; Artificial intelligence

Classification: 0548: Mechanical engineering; 0984: Computer science; 0800: Artificial intelligence

Identifier / keyword: Applied sciences

Number of pages: 209

Publication year: 1991

Degree date: 1991

School code: 0028

Source: DAI-B 53/05, Dissertation Abstracts International

Place of publication: Ann Arbor

Country of publication: United States

Advisor: Agogino, Alice M.

University/institution: University of California, Berkeley

University location: United States -- California

Degree: Ph.D.

Source type: Dissertations &Theses

Language: English

Document type: Dissertation/Thesis

Dissertation/thesis number: 9228887

ProQuest document ID: 303917765

Document URL: https://search.proquest.com/docview/303917765?accountid=8243

Copyright: Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.

Database: ProQuest Dissertations &Theses Global



Analysis of self-organizing neural networks with application to pattern classification
Lo, Zhen-Ping . University of California, Irvine, ProQuest Dissertations Publishing, 1991. 9212431.
ProQuest document link




ABSTRACT
Recent studies have indicated that neural networks can be implemented to solve the pattern recognition problem. The formulation of the pattern classification problem by self-organizing neural networks, specifically investigation of the Kohonen neural networks are presented in this work. The Kohonen Topology Preserving Mapping (TPM) network and the Learning Vector Quantization (LVQ) algorithms are reviewed. A formal analysis of the convergence property and the neighborhood interaction function selection in the topology preserving unsupervised neural network are presented. Furthermore, the derivation and convergence of the LVQ algorithms are investigated. A neural network piecewise linear classifier based on the Kohonen LVQ2 algorithm and the Kohonen TPM network is developed. The neural network classifier is tested on both synthesized and real data sets. The performance of the proposed classifier is compared with other neural network classifiers and classical classifiers.

LINKS
DETAILS

Subject: Electrical engineering; Automotive materials; Systems design; Artificial intelligence

Classification: 0544: Electrical engineering; 0540: Automotive materials; 0790: Systems design; 0800: Artificial intelligence

Identifier / keyword: Applied sciences

Number of pages: 173

Publication year: 1991

Degree date: 1991

School code: 0030

Source: DAI-B 52/12, Dissertation Abstracts International

Place of publication: Ann Arbor

Country of publication: United States

Advisor: Bavarian, Behnam

University/institution: University of California, Irvine

University location: United States -- California

Degree: Ph.D.

Source type: Dissertations &Theses

Language: English

Document type: Dissertation/Thesis

Dissertation/thesis number: 9212431

ProQuest document ID: 303921519

Document URL: https://search.proquest.com/docview/303921519?accountid=8243

Copyright: Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.

Database: ProQuest Dissertations &Theses Global



Symbolic knowledge and neural networks: Insertion, refinement and extraction
Towell, Geoffrey Gilmer . The University of Wisconsin - Madison, ProQuest Dissertations Publishing, 1991. 9209252.
ProQuest document link




ABSTRACT
Explanation-based and empirical learning are two largely complementary methods of machine learning. These approaches to machine learning both have serious problems which preclude their being a general purpose learning method. However, a "hybrid" learning method that combines explanation-based with empirical learning may be able to use the strengths of one learning method to address the weaknesses of the other method. Hence, a system that effectively combines the two approaches to learning can be expected to be superior to either approach in isolation. This thesis describes a hybrid system called K scBANN which is shown to be an effective combination of these two learning methods.
K scBANN (Knowledge-Based Artificial Neural Networks) is a three-part hybrid learning system built on top of "neural" learning techniques. The first part uses a set of approximately-correct rules to determine the structure and initial link weights of an artificial neural network, thereby making the rules accessible for modification by neural learning. The second part of K scBANN modifies the resulting network using essentially standard neural learning techniques. The third part of K scBANN extracts refined rules from trained networks.
K scBANN is evaluated by empirical tests in the domain of molecular biology. Networks created by K scBANN are shown to be superior, in terms of their ability to correctly classify unseen examples, to a wide variety of learning systems as well as techniques proposed by experts in the problems investigated. In addition, empirical tests show that K scBANN is robust to errors in the initial rules and insensitive to problems resulting from the presence of extraneous input features.
The third part of K scBANN, which extracts rules from trained networks, addresses a significant problem in the use of neural networks--understanding what a neural network learns. Empirical tests of the proposed rule-extraction method show that it simplifies understanding of trained networks by reducing the number of: consequents (hidden units), antecedents (weighted links), and possible antecedent weights. Surprisingly, the extracted rules are often more accurate at classifying examples not seen during training than the trained network from which they came.

LINKS
DETAILS

Subject: Computer science; Artificial intelligence

Classification: 0984: Computer science; 0800: Artificial intelligence

Identifier / keyword: Applied sciences

Number of pages: 263

Publication year: 1991

Degree date: 1991

School code: 0262

Source: DAI-B 53/02, Dissertation Abstracts International

Place of publication: Ann Arbor

Country of publication: United States

Advisor: Shavlik, Jude William

University/institution: The University of Wisconsin - Madison

University location: United States -- Wisconsin

Degree: Ph.D.

Source type: Dissertations &Theses

Language: English

Document type: Dissertation/Thesis

Dissertation/thesis number: 9209252

ProQuest document ID: 303927450

Document URL: https://search.proquest.com/docview/303927450?accountid=8243

Copyright: Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.

Database: ProQuest Dissertations &Theses Global



Biomedical signal processing and pattern recognition by artificial neural networks
Xue, Qiuzhen . The University of Wisconsin - Madison, ProQuest Dissertations Publishing, 1991. 9123875.
ProQuest document link




ABSTRACT
We developed an artificial-neural-network-based adaptive filter (ANNADF) for nonlinear biomedical signal filtering and modeling. We addressed the issues of (1) the stability condition, (2) convergence rate, (3) generalization capability for noise elimination, and (4) the sensitivity towards weight error of the ANNADF. We tested the performance of the ANNADF for simulated linear and nonlinear signals and sampled biomedical signals. Based on the ANNADF, we developed an ANN-based adaptive matched filter for QRS detection, and an ANN-based multichannel adaptive filter for evoked potential signal enhancement. All the results were compared with those of linear filters, and the comparison results show than ANN-based filters outperform linear filters for nonlinear biomedical signal processing applications. We also proposed several methods to reduce the e

/ 0 نظر / 60 بازدید