Дисертації з теми "Multi-criteria estimation"

There are three components of error in the ability of land-atmosphere models (e.g., BATS, SiB, etc.) to simulate/predict observed land-surface state variables and output fluxes (e.g. lambdaE, H, Tg, Q, etc.). The first is caused by model structural error associated with simplifications and/or inadequacies in the functional representations of underlying physical processes. The second component is measurement error associated with the input and output data. The third is caused by error in specification of the values of the model parameters. Automatic parameter tuning (model calibration) methods allow minimizing of the parameter error, thereby obtaining an estimate of the remaining error components. This work describes an automatic multi-criteria approach and its use to tune all 27 parameters of the BATS model using data measured in the field. The parameters were adjusted to simultaneously optimize the ability of the model to reproduce observed values of several output fluxes and/or state variables (e.g., latent heat flux, sensible heat flux, ground temperature, etc.). The results indicate that not only does the procedure result in conceptually reasonable and consistent parameter estimates, but the calibrated model is able to provide significant improvement in performance (33% or more reduction in error) over the "un-calibrated" model (i.e., using the BATS default parameter values for the associated region). Substantial improvements of this kind can have important implications for studies that seek to evaluate alternative model structures or to regionalize parameters. To reduce the dimensionality of the optimization problem a multi-criteria extension of the Regionalized Sensitivity Analysis (RSA) has been developed.

The work presented in this thesis targets the analysis and implementation of multi-criteria performance prediction methods for System-on-Chips (SoC).

These new SoC architectures offer the opportunity to integrate complete heterogeneous systems into a single chip and can be used to design battery powered handhelds, security critical systems, consumer electronics devices, etc. However, this variety in terms of application usually comes with a lot of different performance objectives like power consumption, yield, design cost, production cost, silicon area and many others. These performance requirements are often very difficult to meet together so that SoC design usually relies on making the right design choices and finding the best performance compromises.

In parallel with this architectural paradigm shift, new Very Deep Submicron (VDSM) silicon processes have more and more impact on the performances and deeply modify the way a VLSI system is designed even at the first stages of a design flow.

In such a context where many new technological and system related variables enter the game, early exploration of the impact of design choices becomes crucial to estimate the performance of the system to design and reduce its time-to-market.

In this context, this thesis presents:

- A study of state-of-the-art tools and methods used to estimate the performances of VLSI systems and an original classification based on several features and concepts that they use. Based on this comparison, we highlight their weaknesses and lacks to identify new opportunities in performance prediction.

- The definition of new concepts to enable the automatic exploration of large design spaces based on flexible performance criteria and degrees of freedom representing design choices.

- The implementation of a couple of two new tools of our own:

- Nessie, a tool enabling hierarchical representation of an application along with its platform and automatically performs the mapping and the estimation of their performance.

-Yeti, a C++ library enabling the defintion and value estimation of closed-formed expressions and table-based relations. It provides the user with input and model sensitivity analysis capability, simulation scripting, run-time building and automatic plotting of the results. Additionally, Yeti can work in standalone mode to provide the user with an independent framework for model estimation and analysis.

To demonstrate the use and interest of these tools, we provide in this thesis several case studies whose results are discussed and compared with the literature.

Using Yeti, we successfully reproduced the results of a model estimating multi-core computation power and extended them thanks to the representation flexibility of our tool.

We also built several models from the ground up to help the dimensioning of interconnect links and clock frequency optimization.

Thanks to Nessie, we were able to reproduce the NoC power consumption results of an H.264/AVC decoding application running on a multicore platform. These results were then extended to the case of a 3D die stacked architecture and the performance benefits are then discussed.

We end up by highlighting the advantages of our technique and discuss future opportunities for performance prediction tools to explore.
Doctorat en Sciences de l'ingénieur
info:eu-repo/semantics/nonPublished

The work presented in this thesis aims at validating an original multicriteria performances estimation tool, NESSIE, dedicated to the prediction of performances to accelerate the design of electronic embedded systems.

This tool has been developed in a previous thesis to cope with the limitations of existing design tools and offers a new solution to face the growing complexity of the current applications and electronic platforms and the multiple constraints they are subjected to.

More precisely, the goal of the tool is to propose a flexible framework targeting embedded systems in a generic way and enable a fast exploration of the design space based on the estimation of user-defined criteria and a joint hierarchical representation of the application and the platform.

In this context, the purpose of the thesis is to put the original framework NESSIE to the test to analyze if it is indeed useful and able to solve current design problems. Hence, the dissertation presents :

- A study of the State-of-the-Art related to the existing design tools. I propose a classification of these tools and compare them based on typical criteria. This substantial survey completes the State-of-the-Art done in the previous work. This study shows that the NESSIE framework offers solutions to the limitations of these tools.

- The framework of our original mapping tool and its calculation engine. Through this presentation, I highlight the main ingredients of the tool and explain the implemented methodology.

- Two external case studies that have been chosen to validate NESSIE and that are the core of the thesis. These case studies propose two different design problems (a reconfigurable processor, ADRES, applied to a matrix multiplication kernel and a 3D stacking MPSoC problem applied to a video decoder) and show the ability of our tool to target different applications and platforms.

The validation is performed based on the comparison of a multi-criteria estimation of the performances for a significant amount of solutions, between NESSIE and the external design flow. In particular, I discuss the prediction capability of NESSIE and the accuracy of the estimation.

-The study is completed, for each case study, by a quantification of the modeling time and the design time in both flows, in order to analyze the gain achieved by our tool used upstream from the classical tool chain compared to the existing design flow alone.

The results showed that NESSIE is able to predict with a high degree of accuracy the solutions that are the best candidates for the design in the lower design flows. Moreover, in both case studies, modeled respectively at a low and higher abstraction level, I obtained a significant gain in the design time.

However, I also identified limitations that impact the modeling time and could prevent an efficient use of the tool for more complex problems.

To cope with these issues, I end up by proposing several improvements of the framework and give perspectives to further develop the tool.
Doctorat en Sciences de l'ingénieur
info:eu-repo/semantics/nonPublished

Ce travail de thèse doctorale traite l'analyse de la vulnérabilité des systèmes critiques pour la sécurité (par exemple, les centrales nucléaires) dans un cadre qui combine les disciplines de l'analyse des risques et de la prise de décision de multi-critères.La contribution scientifique suit quatre directions: (i) un modèle hiérarchique et quantitative est développé pour caractériser la susceptibilité des systèmes critiques pour la sécurité à plusieurs types de danger, en ayant la vue de `tous risques' sur le problème actuellement émergeant dans le domaine de l'analyse des risques; (ii) l'évaluation quantitative de la vulnérabilité est abordé par un cadre de classification empirique: à cette fin, un modèle, en se fondant sur la Majority Rule Sorting (MR-Sort) Méthode, généralement utilisés dans le domaine de la prise de décision, est construit sur la base d'un ensemble de données (en taille limitée) représentant (a priori connu) des exemples de classification de vulnérabilité; (iii) trois approches différentes (à savoir, une model-retrieval-based méthode, la méthode Bootstrap et la technique de validation croisée leave-one-out) sont élaborées et appliquées pour fournir une évaluation quantitative de la performance du modèle de classification (en termes de précision et de confiance dans les classifications), ce qui représente l'incertitude introduite dans l'analyse par la construction empirique du modèle de la vulnérabilité; (iv) basé sur des modèles développés, un problème de classification inverse est résolu à identifier un ensemble de mesures de protection qui réduisent efficacement le niveau de vulnérabilité du système critique à l’étude. Deux approches sont développées dans cet objectif: le premier est basé sur un nouvel indicateur de sensibilité, ce dernier sur l'optimisation.Les applications sur des études de cas fictifs et réels dans le domaine des risques de centrales nucléaires démontrent l'efficacité de la méthode proposée
This Ph. D. work addresses the vulnerability analysis of safety-critical systems (e.g., nuclear power plants) within a framework that combines the disciplines of risk analysis and multi-criteria decision-making. The scientific contribution follows four directions: (i) a quantitative hierarchical model is developed to characterize the susceptibility of safety-critical systems to multiple types of hazard, within the needed `all-hazard' view of the problem currently emerging in the risk analysis field; (ii) the quantitative assessment of vulnerability is tackled by an empirical classification framework: to this aim, a model, relying on the Majority Rule Sorting (MR-Sort) Method, typically used in the decision analysis field, is built on the basis of a (limited-size) set of data representing (a priori-known) vulnerability classification examples; (iii) three different approaches (namely, a model-retrieval-based method, the Bootstrap method and the leave-one-out cross-validation technique) are developed and applied to provide a quantitative assessment of the performance of the classification model (in terms of accuracy and confidence in the assignments), accounting for the uncertainty introduced into the analysis by the empirical construction of the vulnerability model; (iv) on the basis of the models developed, an inverse classification problem is solved to identify a set of protective actions which effectively reduce the level of vulnerability of the critical system under consideration. Two approaches are developed to this aim: the former is based on a novel sensitivity indicator, the latter on optimization.Applications on fictitious and real case studies in the nuclear power plant risk field demonstrate the effectiveness of the proposed methodology

L'Union européenne (UE) a établi la prospective politique pour atteindre ses objectifs énergétiques fondamentaux pour le développent durable, la compétitivité et la sécurité énergétique, en réduisant les émissions de gaz à effet de serre par le biais d'une augmentation de la part des énergies renouvelables dans la consommation d'énergie et par l'amélioration de l'efficacité énergétique. Les principales questions sur l'utilisation à grande échelle des EnR sont liées au dimensionnement des systèmes, le choix parmi une grande variété de solutions vis-à-vis de plusieurs critères, et enfin le contrôle de ces sources. Dans un proche avenir, de plus en plus les EnR vont cohabiter avec les sources d'énergie fossiles et la recherche doit être orientée vers des solutions qui sont efficaces du point de vue énergétique, économiquement viable et respectueuses de l'environnement. Dans cette thèse, les travaux de recherche établissent une démarche en vue de proposer des solutions qui pourraient être les réponses aux deux premières problématiques que sont le dimensionnement mais surtout la seconde, qui est le choix des systèmes énergétiques les mieux adaptés par rapport a un nombre donné de critères. La première partie de la thèse traite des questions liées à l'estimation des besoins pour le chauffage, l'eau chaude sanitaire et l'énergie électrique. Cette estimation a été abordée par le développement de modèles polynomiaux de régression. La deuxième partie de la thèse examine les techniques de modélisation des systèmes afin d'obtenir les profils de l'approvisionnement en énergie renouvelable : ces quantités d'énergie de l'offre sont nécessaires dans d'analyse décisionnelle. Dans la troisième partie, une méthode d'aide à la décision multicritères (ELECTRE III) est décrite, puis appliquée à un exemple. La technique consiste à utiliser les règles obtenues à partir des modèles précédents et à réaliser un classement de solutions possibles envisagées. L'analyse multicritères fournit ainsi une technique intéressante d'aide dans le processus de décision et est capable de justifier les choix et de surclasser les alternatives dans le secteur des énergies renouvelables. Enfin, les théories, les algorithmes et les modèles qui ont été décrits dans ce travail ont été intégrées dans un outil d'aide à la décision que nous avons développé visant spécifiquement l'intégration des technologies à énergie renouvelable par une analyse multicritères et de faire rapidement des études paramétriques sur les consommations d'énergie des bâtiments. Dans la dernière partie de la thèse une analyse complète d'une étude de cas (maison Mozart) est réalisée avec la mise en application des méthodes proposées
European Union (EU) has agreed a forward-looking political agenda to achieve its core energy objectives of sustainability, competitiveness and security of supply, by reducing greenhouse gas emissions through an increase of the share of renewable in the energy consumption and by improving energy efficiency. The main issues of renewable energy sources large scale use are related to the sizing of the systems, the choice among a large variety of alternatives face to a certain number of criteria, and finally the control of these sources. In the near future, more and more the RES will cohabit with fossil energy source systems and research has to be pointed towards solutions that are energy efficiently, economical viable and environmental friendly. In this thesis, the research work is focus on finding and proposing solutions that could be the answers for the first two main issues presented previous, especially on the second issue which is the choice of systems face to several criteria. The first part treats the issues related to the heating, domestic hot water and electricity energy demand assessment, from the estimation to the impact factors. At this level it is described a new methodology developed to estimate the heating demand of residential buildings in temperate climate by using polynomial regression models obtained from a database of values from dynamic simulations. The second part examines the modeling techniques to obtain the renewable energy supply profiles which are further used in the multicriteria decision analysis. In the third part, a multicriteria decision support methodology concept is de-scribed (ELECTRE III) and then applied for an example. The decision support algorithm has its bases on the developed models and realizes the outranking of the possible. Finally, the theories, algorithms and models that have described in the thesis have been encapsulated in the development of a decision support-tool specifically aimed to aid strategic decisions regarding renewable technology integration and making quick parametric studies on the building energy consumptions. In the last part of the thesis a complete analysis of a study case (Mozart dwelling) is realized with application of the proposed methodologies

Fire plays an important role in forest ecosystem management depending on the dual character of it. It should be managed and supervised effectively. In this particular study, the study area was located in the north part of Gävleborg County in Sweden, which is in a high- latitude region. Seven factors, divided into natural factors and human caused factors, were extracted from digital elevation model (DEM), classified land use map and feature shape files provided by National Land Survey of Sweden (Lantmäteriet). Two different weighting schemes for the factors were determined by the Analytic Hierarchy Process (AHP) method. With the help of ArcGIS 9.3 and Erdas 9.3, two classified result maps were obtained, where forest fire risk ranks were shown as five classes, very low, low, moderate, high and very high. The 43 fire incidents in the year 2007 and 2008 recorded by Global Fire Management System were used to evaluate the results. The results show that the higher rank the region is, the larger is the probability for forest fire risk and higher the risk to spread the fire. Furthermore, according to the occurrence time of the fire incidents, the period of time from end of May to beginning of June was generalized as a dangerous period for forest fire risk in this study area. After analyzing and discussing, even if there might be some uncertainties caused by variable selection, resolution problem and weighting schemes, the results were generally reliable.

Рассматривается научно-техническая проблема синтеза интеллектуальных систем поддержки принятия решений при обработке биомедицинских сигналов и изображений. В работе используются методы цифровой обработки сигналов и изображений, многокритериальные оценки, методы выделения новых знаний с целью повышения качества принятия решений в условиях априорной неопределенности.
Scientific and technical problem of synthesis of intelligent decision support systems in biomedical signals and images processing is considered. Methods of digital signal and image processing, multi-criteria estimation, methods for detection of new knowledge to improve the quality of decision making under prior uncertainty are used in the work.

Les récentes évolutions des technologies de l'information et de la communication font qu'il est aujourd'hui facile de consulter des catalogues de morceaux de musique conséquents. De nouvelles représentations et de nouveaux algorithmes doivent de ce fait être développés afin de disposer d'une vision représentative de ces catalogues et de naviguer avec agilité dans leurs contenus. Ceci nécessite une caractérisation efficace des morceaux de musique par l'intermédiaire de descriptions macroscopiques pertinentes. Dans cette thèse, nous nous focalisons sur l'estimation de la structure des morceaux de musique : il s'agit de produire pour chaque morceau une description de son organisation par une séquence de quelques dizaines de segments structurels, définis par leurs frontières (un instant de début et un instant de fin) et par une étiquette représentant leur contenu sonore.La notion de structure musicale peut correspondre à de multiples acceptions selon les propriétés musicales choisies et l'échelle temporelle considérée. Nous introduisons le concept de structure "sémiotique" qui permet de définir une méthodologie d'annotation couvrant un vaste ensemble de styles musicaux. La détermination des segments structurels est fondée sur l'analyse des similarités entre segments au sein du morceau, sur la cohérence de leur organisation interne (modèle "système-contraste") et sur les relations contextuelles qu'ils entretiennent les uns avec les autres. Un corpus de 383 morceaux a été annoté selon cette méthodologie et mis à disposition de la communauté scientifique.En termes de contributions algorithmiques, cette thèse se concentre en premier lieu sur l'estimation des frontières structurelles, en formulant le processus de segmentation comme l'optimisation d'un coût composé de deux termes~: le premier correspond à la caractérisation des segments structurels par des critères audio et le second reflète la régularité de la structure obtenue en référence à une "pulsation structurelle". Dans le cadre de cette formulation, nous comparons plusieurs contraintes de régularité et nous étudions la combinaison de critères audio par fusion. L'estimation des étiquettes structurelles est pour sa part abordée sous l'angle d'un processus de sélection d'automates à états finis : nous proposons un critère auto-adaptatif de sélection de modèles probabilistes que nous appliquons à une description du contenu tonal. Nous présentons également une méthode d'étiquetage des segments dérivée du modèle système-contraste.Nous évaluons différents systèmes d'estimation automatique de structure musicale basés sur ces approches dans le cadre de campagnes d'évaluation nationales et internationales (Quaero, MIREX), et nous complétons cette étude par quelques éléments de diagnostic additionnels.

Information retrieval is much more challenging than traditional small document collection retrieval. The main difference is the importance of correlations between related concepts in complex data structures. These structures have been studied by several information retrieval systems. This research began by performing a comprehensive review and comparison of several techniques of matrix dimensionality estimation and their respective effects on enhancing retrieval performance using singular value decomposition and latent semantic analysis. Two novel techniques have been introduced in this research to enhance intrinsic dimensionality estimation, the Multi-criteria Decision Weighted model to estimate matrix intrinsic dimensionality for large document collections and the Average Standard Estimator (ASE) for estimating data intrinsic dimensionality based on the singular value decomposition (SVD). ASE estimates the level of significance for singular values resulting from the singular value decomposition. ASE assumes that those variables with deep relations have sufficient correlation and that only those relationships with high singular values are significant and should be maintained. Experimental results over all possible dimensions indicated that ASE improved matrix intrinsic dimensionality estimation by including the effect of both singular values magnitude of decrease and random noise distracters. Analysis based on selected performance measures indicates that for each document collection there is a region of lower dimensionalities associated with improved retrieval performance. However, there was clear disagreement between the various performance measures on the model associated with best performance. The introduction of the multi-weighted model and Analytical Hierarchy Processing (AHP) analysis helped in ranking dimensionality estimation techniques and facilitates satisfying overall model goals by leveraging contradicting constrains and satisfying information retrieval priorities. ASE provided the best estimate for MEDLINE intrinsic dimensionality among all other dimensionality estimation techniques, and further, ASE improved precision and relative relevance by 10.2% and 7.4% respectively. AHP analysis indicates that ASE and the weighted model ranked the best among other methods with 30.3% and 20.3% in satisfying overall model goals in MEDLINE and 22.6% and 25.1% for CRANFIELD. The weighted model improved MEDLINE relative relevance by 4.4%, while the scree plot, weighted model, and ASE provided better estimation of data intrinsic dimensionality for CRANFIELD collection than Kaiser-Guttman and Percentage of variance. ASE dimensionality estimation technique provided a better estimation of CISI intrinsic dimensionality than all other tested methods since all methods except ASE tend to underestimate CISI document collection intrinsic dimensionality. ASE improved CISI average relative relevance and average search length by 28.4% and 22.0% respectively. This research provided evidence supporting a system using a weighted multi-criteria performance evaluation technique resulting in better overall performance than a single criteria ranking model. Thus, the weighted multi-criteria model with dimensionality reduction provides a more efficient implementation for information retrieval than using a full rank model.
Ph.D.
Department of Industrial Engineering and Management Systems
Engineering and Computer Science
Industrial Engineering PhD

Magistro studijų baigiamasis darbas, 67 puslapiai, 20 paveikslų, 9 lentelės, 50 literatūros šaltinių, 2 priedai, lietuvių kalba. Tyrimo objektas – apskaitos konsultavimo paslaugų kokybė. Tyrimo tikslas – parengti apskaitos konsultavimo paslaugų kokybės daugiakriterinio vertinimo modelį. Tyrimo uždaviniai: 1) išanalizuoti apskaitos, apskaitos konsultavimo paslaugų, kokybės sampratas; 2) išanalizuoti konsultavimo paslaugų kokybės vertinimo modelius; 3) parengti apskaitos konsultavimo paslaugų kokybės vertinimo kriterijų sistemą ir daugiakriterinį kokybės vertinimo modelį; 4) įvertinti apskaitos konsultavimo paslaugų kokybę Kauno, Jurbarko ir Radviliškio rajonų savivaldybėse; 5) pateikti apskaitos konsultavimo paslaugų kokybės vertinimo proceso tobulinimo galimybes. Tyrimo laikotarpis: 2000 m. – 2007 m. Tyrimo metodai: monografinis analizės ir sintezės, palyginimo, anketinė apklausa, loginio mąstymo, grafinio vaizdavimo ir kiti.
Final paper of Master studies, 67 pages, 20 pictures, 9 tables, 50 literature sources, 2 appendixes, in Lithuanian. Research object – quality of accounting consultation service. Research aim – to prepare multiobjective valuation model of accounting consultation service quality. Research tasks: 1) to sift conceptions of accounting, accounting consultation service, quality; 2) to analyse consultation service quality estimation models; 3) to prepare estimation criteria system for accounting consultation service quality and model of multi-criteria quality valuation; 4) to estimate quality of accounting consultation service in Kaunas, Jurbarkas and Radviliskis districts’ municipalities; 5) to introduce recommendations how to improve accounting consultation service quality. Time of exploration: 2000 – 2007 year. Research methods: monographic analysis and synthesis, comparison, questionnaire, logic way of thinking, graphical representation and others.

This thesis describes the issue of selecting of the company's clients Fragile media s. r. o. The company looks at the financial benefits and time. The target of this thesis is to determine what kind of clients brings to Fragile media s. r. o. the greatest benefit. This issue I will discuss in the practical part of this thesis by using multi-criteria decision making. In next step I will deal with what effect will have other methods of estimating weights of the criteria for the results.

碩士
國立清華大學
電機工程學系
89
Tugnait, Chi and Chen proposed multi-input multi-output inverse filter criteria (MIMO-IFC) using higher-order statistics (HOS) for blind deconvolution of MIMO linear time-invariant (LTI) systems. Chi and Chen also proposed a property of nonlinear relationship between the optimum inverse filter associated with MIMO-IFC and the MIMO minimum mean square error (MIMO-MMSE) equalizer. In this paper, based on the nonlinear relationship, an iterative single-input multi-output (SIMO) blind channel estimation (BCE) algorithm and a time delay estimation (TDE) algorithm using HOS are proposed with a given set of non-Gaussian measurements. The former is a fast FFT-based algorithm for joint estimation of magnitude response and phase response of SIMO systems and the latter estimates P-1 time delays from the phase information of single-input P-output (P>=2) system leading to a closed-form solution of time delay estimates using the optimum inverse filter associated with MIMO-IFC. Some simulation results are presented to support the efficacy of the proposed SIMO BCE and TDE algorithms.

Bike share system (BSS) has received increasing attention in research for its potential economic and environmental benefits. However, some research has pointed out the negative sustainability impacts of BSS from rebalancing activity, due to its greenhouse gas (GHG) emissions and additional vehicle travels. Additionally, bike and station manufacturing also bring considerable emissions to the system. Therefore, it is important to analyze the current rebalancing efficiency and sustainability of BSSs, and to assist the BSS operators in optimizing the BSS design. Existing studies lack tools to estimate the real-world rebalancing activities and vehicle usage for system sustainability evaluation and improvements. To address this gap, this research first proposed a framework to estimate rebalancing activities and applied a clustering-based method to estimate the rebalancing vehicle use. Applying the framework to the BSSs in Chicago, Boston, and Los Angeles, this study estimated the rebalancing operation and compared the rebalancing efficiencies among the three systems. The analysis results show that 1) only a small proportion of stations and bikes were involved in the daily rebalancing activities; 2) most rebalancing activities were operated during the daytime, while the overnight rebalancing was limited; 3) the system scale, trip demand, and station types are critical for the rebalancing efficiency; and 4) reducing the rebalancing activities at self-rebalance stations could help to improve the rebalancing efficiency and benefits system sustainability. Additionally, the sustainability performance (e.g., carbon emissions) of BSS is not only decided by the rebalance, but also the manufacturing of bikes and stations. It is important to consider all these factors when optimizing a BSS. The existing literature on system improvement for the BSSs lacks an integrated view, and a well-designed integrated model for current BSS improvement is needed. The second part of this thesis built a simulation-based optimization model and generated 2400 scenarios for evaluation. This model aims to minimize the expansion investment, rebalancing mileage, and maximize the system demand and service rate. A Weight Sum Model is applied to solve the multi-criteria decision analysis. The model results show that the best system improvement is to build a new station with a small capacity and initial bikes. The investment and location impacts are discussed to find the tradeoff among expansion strategies. A sensitivity analysis is conducted to evaluate how different weight combinations (refer to different preferences in decision making) impact the preferred station configuration (docks and bikes) and new station locations.