The National Authority for Scientific Research




Скачать 216.78 Kb.
НазваниеThe National Authority for Scientific Research
страница1/4
Дата конвертации29.10.2012
Размер216.78 Kb.
ТипДокументы
  1   2   3   4

ROMANIA


Ministry of Education, Research and Youth


The National Authority for Scientific Research









     PN–II–ID–PCE–2008–2


FUNDING APLICATION FOR EXPLORATORY RESEARCH PROJECTS

­


1. Personal data of the project manager:

1.1. Surname:

ISAR

1.2. First name:

ALEXANDRU

1.3. Year of birth:

1957

1.4. Didactic and/or scientific title:

(Select)

1.5. PhD since the year:

1993

1.6 Doctorate coordinator:

(Select)

    1. Number of candidates for

doctor’s degree:

3



2. Host institution:

2.1. The name of the institution:

UNIVERSITATEA POLITEHNICA TIMISOARA [fill in the institution name]

2.2 Faculty/ Department:

ELECTRONICS AND TELECOMMUNICATIONS/COMMUNICATIONS

2.3. Position:

PROFESSOR

2.4 Address:

2 BD. V. PARVAN,

2.5 Telephone:

40 256 403307

2.6 Fax:

40 256 403295

2.7. E-Mail:

alexandru.isar@etc.upt.ro



3. Title of the project: (Max. 200 characters)

USING WAVELETS THEORY FOR DECISION MAKING



4. Key words (max. 5 terms ):

1

WAVELET FUNCTIONS

2

DECISION MAKING

3

DENOISING

4

SEGMENTATION

5

CLASSIFICATION







5. Project duration ( 3 years):


6. Project summary: (Max. 2000 characters)

MAKING DECISIONS IS A BRANCH OF ARTIFICIAL INTELLIGENCE THAT IS MORE AND MORE USED IN COMPLEX APPLICATIONS LIKE MEDICINE (USING A DIAGNOSTIC A TREATMENT DECISION IS MADE), GEOLOGY (USING IMAGES OF A REGION SOME HYPOTHESES REGARDING THE UNDERGROUND COMPOSITION AND SOME DECISION ABOUT EXTRACTION ARE MADE) OR COMMUNICATIONS (USING INFORMATION ABOUT THE FUNCTIONING OF EACH ELEMENT OF A COMMUNICATION NETWORK SOME DECISIONS ABOUT THE RESOURCES ALLOCATION ARE MADE, FOR EXAMPLE OF THE FREQUENCY BANDWIDTHS). ACCORDING TO BOB COLWELL, ANY MACHINE CAN HAVE ARTIFICIAL INTELLIGENCE. THIS MUST BE DEVELOPED ON THE BASIS OF UNDERSTANDING AND IMITATION OF THE HUMAN BRAIN. THE INTELLIGENCE RESULTS FROM THE ACTION OF A LARGE GROUP OF SPECIALIZED NEURONS THAT USE A WORLD MODEL BASED ON MEMORY TO MAKE A CONTINUOUS SERIES OF PREDICTIONS OF FUTURE EVENTS. THE NEURAL NETWORKS OF THE CORTEX MUST BE INTERPRETED LIKE A DISTRIBUTED MEMORY OF PATTERN SEQUENCES STOKED IN AN INVARIANT FORM, HIERARCHICALLY ARRANGED, ACCESSED IN AN ASSOCIATIVE FASHION. BETWEEN THE NEURAL NETWORK APPLICATIONS ALREADY KNOWN WE CAN FIND APPLICATIONS IN DECISION MAKING FOR MEDICINE, GEOLOGY AND COMMUNICATIONS. TO MAKE A CORRECT DECISION, THE DECIDER MUST HAVE THE INFORMATION IN AN APPROPRIATE FORM. THIS IS THE REASON WHY FREQUENTLY ARE USED ALTERNATIVE REPRESENTATIONS OF INFORMATION. A VERY INTERESTING REPRESENTATION IS IN THIS RESPECT THE WAVELET DECOMPOSITION. IN THIS PROJECT WE WANT TO ASSOCIATE THE WAVELETS THEORY WITH THE NEURAL NETWORK THEORY TO SOLVE PROBLEMS OF DECISIONS MAKING IN MEDICINE, IN GEOLOGY AND IN COMMUNICATIONS. TO DO THIS WE ASSOCIATED THE COMPETENCES OF TWO SENIOR RESEARCHERS IN THE FIELD OF NEURAL NETWORKS WITH THE COMPETENCES IN THE WAVELET THEORY OF ALL SIX MEMBERS OF OUR RESEARCH TEAM.



7. Project presentation:

[Please fill in max. 10 pages in ANNEX 1]



8. Project management:

[Please fill In ANNEX 2]



9. Budget (eligible cost)*:

CRT. NO.



NAME OF THE BUDGET CATEGORY

VALUE

2008***

(euro)

VALUE

2009***

(euro)

VALUE

2010***

(euro)

VALUE

2011***

(euro)

TOTAL VALUE

(euro)

1.

STAFF EXPENSES** - max. 60%

including state tax and other contribution

3450

15000

15000

11850

45300

2.

INDIRECT EXPENSES (overheads)

2250

9000

9000

6750

27000

3.

MOBILITIES

(participation in prestigious scientificevents / documentation; research stages in contry or abroad)

13100

49600

49600

44900

157200

4.

LOGISTIC COSTS for carrying on the project

(research infrastructure, costs for materials, dissemination etc.)

3700

16400

16400

4000

40500




TOTAL

22500

90000

90000

67500

270000


* Structure of budget must be in accordance with the HG 1579/2002

** Staff expanses must be calculated in acordance with the size of research team and HG 475/2007

*** 2008 – 3 month, 2009 – 12 month; 2010 – 12 month; 2011 – 9 month.



  1. Project manager is full time employed in the host institution




(Select)







IT IS CERTIFIED HEREBY THE LEGALITY AND CORRECTNESS

OF THE DATA INCLUDED IN THE PRESENT FINANCING REQUEST


DATE: FEBRUARY, 24, 2007


RECTOR/MANAGER,

Surname, first name:ROBU Nicolae

Signature:

Seal


ACCOUNTING MANAGER/CHIEF ACCOUNTANT

Surname, first name:MICLEA Florian

Signature:




PROJECT MANAGER,

Surname, first name:ISAR ALEXANDRU

Signature:





ANNEX 1

7. Project presentation: (Max. 10 pages)


7.1. Importance and relevance of the scientific content

The presentation creates the research referential; it will demonstrate the project manager’s

degree of information documentation

Making decisions is a branch of Artificial Intelligence, AI, that is used more and more frequently in complex applications like medicine (on the basis of a diagnostic a treatment decision is done), geology (using images of a region some hypotheses concerning the underground content and some extraction decisions are made), financial analysis (on the basis of temporal series some financial politic decisions are made) or communications (using information about operation of each element of a communications network some decisions about the resources allocation are made, for example the frequency bandwidths are allocated). Like Bob Colwell, former manager at Intel, writes in his article: Machine Intelligence Meets Neuroscience, published in the journal Computer, of the IEEE computer science section in January 2005, ([1]), any machine, regarded like a collection of materials, combined to work together, can have intelligence. This kind of intelligence is named AI because in their flight the planes don’t mimic the birds and the propulsion of boats is different of the propulsion of fishes. A high enthusiasm regarding the future of the AI was observed 25 years ago, which unfortunately is not so intensive today. In October 2004 was published the Jeff Hawkins’ book, entitled On Intelligence ([2]). The central idea of this book is that the development of the AI must follows the understanding and the imitation of the human brain functioning. Hawkins believes that the intelligence results from the action of a massive group of specialized neurons that use a world model based on memory to make a continuous series of predictions of future events. He believes that the human brain comportment has 3 very important aspects: the human brain works with feedback connections, it uses temporal sequences input strings; it makes hierarchies of cortex neural networks, NN, used for the realization of specialized functions. Hawkins suggests that the cortex NN must be interpreted like a distributed memory of pattern sequences stoked in an invariant form, arranged in a hierarchy, associatively accessed. Colwell agrees these conclusions and highlights the principal role of the NN in AI. This role is detailed in: [3]-Michael A. Arbib, editor, The Hanbook of Brain Theory and Neural Networks, MIT, 2003, ISBN 0-262-011967-2. This book has more than 1000 pages. The first two chapters are written by the editor to guide the reader in the lecture of the third chapter that represents a collection of the best papers written recently in the field of NN. We can find here NN applications for making decisions in medicine, geology, financial analysis and in communications. The most popular applications of NN (for example their use to implement genetic algorithms) are presented in the book: [4]-Juan R. Rabunal, Julian Dorado, Artificial Neural Networks in Real-Life Applications, Idea Group Inc., 2006. The signal processing, SP, applications of NN are presented in the book: [5]-Yu Hen Hu, Jenq-Neng Hwang, editors, Handbook of Neural Network Signal Processing, CRC Press, 2001, ISBN 0_8493_2359_2. The decisions making can be assisted by expert systems, very popular in the AI literature around 1980. Such an expert system is presented in the paper: [6]-J. K. Tsotos, H. D. Covvey, J. Mylopoulos, P. McLaughlin, The Role of Symbolic Processing in the Computer Evaluation of Left Ventricular Wall Motion: The ALVEN System, Medical Engineering & Physics, Volume 21, Issue 2, Pages 73-85. To make a correct decision, the decider must have the existing information in an appropriate form. This is the reason why frequently alternative forms of information are used. For example we can apply a transformation to the signal that carries the information. This can be the Fourier transform (for example for the JPEG compression), the wavelet transform (for example for the JPEG-2000 compression or for denoising) or another time-frequency representation (for example for the instantaneous frequency estimation). Generally, these transforms are associated with corresponding series decompositions. So we can speak about Fourier series or wavelet series. A very important decomposition for the aid at decisions making is the singular value decomposition, SVD. Sometimes, different information comes from different sources. The signals carrying this information are mixed and affected by noise. In this case the decider must have access at the original signals. So, the received signals must be processed to diminish the perturbing noise and then separated. In the case when the original signals were independent, the reduction of the noise and the separation can be done by Independent Component Analysis, ICA. A similar analysis is the Principal Component Analysis, PCA. If only the identification of the independent components is required then the Karhunen-Loeve transform can be used. The block diagram of an acquisition system dedicated to the aid of decisions making is composed by the acquisition system itself by a denoising system, by a segmentation block (that highlights the signal’s regions having similar characteristics) and by a classification block. The phenomenon used for decision making has two states, the normal state (the patient has a good health, the image of the region don’t indicates the presence in the underground of interesting resources, the temporal series keeps its variation sense, the communications network works good) for which the last decision already made works and the exceptional state (the health of the patient is perturbed, the image of the region indicates the presence in the underground of some interesting resources, the sense of variation of the temporal series changed, the communications network is saturated) when new decisions must be made. So, the principal task of the system for aid at decisions making is to identify the exceptional events. One of the transforms that highlights very good these events is the wavelet transform, WT, due its high capacity to separate the details of a signal. This is the reason for which we want to associate in this project the wavelet theory with the NN theory, to solve problems in the field of the aid for decisions making in medicine, geology and communications. The association of these theories is recommended in the following article: [7]-D. A. Karras, S. A. Karkanis, B. G. Mertzios, Information Systems based on neural network and wavelet methods with application to decision making, modeling and prediction tasks, Kybernetes, vol. 27, pp.224-236, 1998. In conformity with this paper, the fields of interest for the application of SP and image processing, IP, are: the decisions making, the prediction and the modeling. In fact the detection and the prediction of anomalies and discontinuities is required. To do that, the wavelets and the NN can be associated. The wavelets offer an adequate time-frequency representation, that permits the anomalies’ localization (hence their detection) and the NN offer a very good solution for classification and approximation. The originality of the solution proposed in this paper lies in the fusion of the information obtained from the WT of a signal with the information carried by the signal itself. This fusion is realized by a NN. Another example for this association is the article: [8]-Ibrahim Turkoglu, Ahmet Arslan, Erdogan Ilkay, A wavelet neural network for the detection of heart valve diseases, Computers in Biology and Medicine, Volume 33, Issue 4, page 319. Here is proposed an expert system for the pattern recognition applied for the diagnosis of cardiac illness. The system is composed by a block for parameters extraction realized with wavelets and by a classifier realized using a perceptron. The applications of NN in communications are described in the following books or articles:

[9]-David W. Corne, Martin J. Oates, George D. Smith, editors, Telecommunications Optimization: Heursitic and Adaptive Techniques, John Wiley&Sons, 2000, ISBN: 0_471_98855_3,

[10]-Murat Husnu Sazli, Can Isik, Neural network implementation of the BCJR algorithm, Digital Signal Processing, Elsevier, 2006, available online at www.sciencedirect.com,

[11]-Bayan S. Sharif, Oliver R. Hinton, T. C. Chuah, Improved Multiuser Detection for Non Gaussian Channels Using Artificial Neural Networks and Turbo Coding, ISCC 2004: 610-614.

Some applications of NN in financial analysis are presented in the following paper:

[12]-Youshen Xia, Henry Leung, Nan Xie, Eloi Bosse, A New Regression Estimator With Neural Network Realization, IEEE Transactions on Signal Processing, vol. 53, no. 2, February 2005.

The notions of NN, learning, SVD and ICA are associated in the book:

[13]-Andrzej Cichocki, Shun-ichi Amari, Adaptive Blind Signal and Image Processing. Learning Algorithms and Applications, John Wiley & Sons, 2002.

One of the recently sustained PhD Thesis in the field of ICA is: [14]-Jan Eriksson, Contributions to theory and algorithms of Independent Component Analysis and Signal Separation, Helsinki University of Technology, August 2004. In this thesis is described the most important application of ICA, the blind identification of sources of independent signals using combinations of affected by noise realizations of those signals. Also, the importance of statistical methods for the appreciation of the degree of independence is highlighted. The optimization algorithms generally used in ICA are also presented. These algorithms minimize some cost functions. The NN can be used in ICA. A way is presented in the paper: [15]-Mark Girolami, Colin Fyfe: Stochastic ICA Contrast Maximisation Using Oja's Nonlinear PCA Algorithm. Int. J. Neural Syst. 8(5-6): 661-678 (1997). The class of mixing models and of sources distribution for real systems implementing ICA can be enlarged using the methods described in the article: [16]-Jan Eriksson, Visa Koivunen, Identifiability, Separability and Uniqueness of Linear ICA Models, IEEE Signal Processing Letters, vol. 11, no. 7, July 2004. A way to assure the independence of the components identified is the whitening of those signals. A new whitening method is proposed in the paper: [17]-Aiyou Chen, Peter J. Bikel, Consistent Independent Component Analysis and Prewhitening, IEEE Transactions on Signal Processing, vol. 53, no. 10, October 2005. The whitening can be performed with wavelets also. The whitening effect is the decorrelation. This can be implemented by dimensions reduction. This method is described in the paper: [18]-Kun Zhang, Lai-Wan Chan, Dimension Reduction as a Deflation Method in ICA, IEEE Signal Processing Letters, vol. 13, no. 1, January, 2006.

Some special cost functions, based on wavelets, are proposed in the article: [19]-Pascal Barbedor, Independent Component Analysis by Wavelets, arXiv:math.ST/0506607v1, 29 June 2005. In the following some applications of ICA in the field of our proposition are presented.

In medicine:

[20]-John. L. Semmlow, Biosignal and Biomedical Image Processing, Marcel Dekker, 2004, ISBN: 0_8247_4803_4,

[21]-Leor Shoker, Saeid Sanei, Jonathon Chambers, Artifact Removal From Electroencephalograms Using a Hybrid BSS-SVM Algorithm, IEEE Signal Processing Letters, vol. 12, no. 10, October 2005,

and in communications – for the channel identification and equalization and for audio communications.

Finally, some applications of wavelets in the fields of interest of this grant are presented.

In geology:

[22]-H. J. Grubb, A. T. Walden, Characterizing seismic time series using the discrete wavelet transform, Geophysical Prospecting, 1997, 45, 183-205,

[23]-D. S. G. Pollok, A Handbook of Time-Series Analysis, Signal Processing and Dynamics, Academic Press, 1999, ISBN: 0_12_560990_6,

[24]-Wenkai Lu, Non-minimum-phase wavelet estimation using second-and third order moments, Geophysical Prospecting, 2005, 53, 149-158.

In communications:

[25]-Patrice Abry, Darryl Veitch, Patrick Flandrin, Long-range Dependence: Revisiting Aggregation with Wavelets, Journal of Time Series Analysis, Volume 19, Number 3, May 1998 , pp. 253-266(14), The aggregation is a natural way to analyze signals having parameters with long dependencies. In this paper is explained how the aggregation can be implemented in the WT domain and how can be applied this procedure for data communication networks.

In statistical signal processing for estimation and classification:

[26]-Felix Abramovich, Umberto Amato, Claudia Angelini, On optimality of Bayesian Wavelet Estimators, Scandinavian Journal of Statistics, vol. 31, 217-234, 2004,

[27]-Keinoosuke Fukunaga, Statistical Pattern Recognition, Academic Press, 1990, ISBN 0-12-269851-7,

[28]-Marina Vannucci, Naijun Sha, Philip J. Brown, NIR and mass spectra classification: Bayesian methods for wavelet-based feature selection, Chemometrics and Intelligent Laboratory Systems, Elsevier, Volume 77, Issues 1-2, 28 May 2005, Pages 139-148.


  1   2   3   4

Добавить в свой блог или на сайт

Похожие:

The National Authority for Scientific Research iconIn this study, I investigated how participating in a scientific research project changed science teachers’ views of scientific inquiry and the nature of

The National Authority for Scientific Research iconThe National Capital Authority

The National Authority for Scientific Research icon2012 National Capital Authority Public Forum

The National Authority for Scientific Research iconThere is no national science just as there is no national multiplication table; what is national is no longer science
А. Kozhevnikova, Assoc. Prof of the Department of English for Humanities (Samara State University), Member of Board of Experts for...

The National Authority for Scientific Research iconSenior Research Associate, 1998-2000 National Research Council, Washington, dc 20418, usa

The National Authority for Scientific Research iconSponsors: Office of Naval Research Air Force Office of Scientific Research ursi (International Commissions g and H) Cooperating Agencies

The National Authority for Scientific Research iconResearch Officer, Pindi Pindi, The National Research Centre for Indigenous Children, Families and Communities, Western Australia

The National Authority for Scientific Research iconAry 505 Scientific Research

The National Authority for Scientific Research iconNational Center for Nuclear Safety and Radiation Control, Atomic Energy Authority, P. O. Box 7551, Nasr City, Cairo 11762, Egypt. Email; Khater

The National Authority for Scientific Research iconScientific Upgrades at the High Flux Isotope Reactor at Oak Ridge National Laboratory


Разместите кнопку на своём сайте:
lib.convdocs.org


База данных защищена авторским правом ©lib.convdocs.org 2012
обратиться к администрации
lib.convdocs.org
Главная страница