<p>This dissertation combines stakeholder and analytical intelligence for consensus decision-making via an interactive optimization process. This dissertation outlines techniques for developing user models of subjective criteria of human stakeholders for an environmental decision support system called WRESTORE. The dissertation compares several user modeling techniques and develops methods for incorporating such user models selectively for interactive optimization, combining multiple objective and subjective criteria. </p><p>This dissertation describes additional functionality for our watershed planning system, called WRESTORE (Watershed REstoration Using Spatio-Temporal Optimization of REsources) (http://wrestore.iupui.edu). Techniques for performing the interactive optimization process in the presence of limited data are described. This work adds a user modeling component that develops a computational model of a stakeholder’s preferences and then integrates the user model component into the decision support system. <br></p><p>Our system is one of many decision support systems and is dependent upon stake- holder interaction. The user modeling component within the system utilizes deep learning, which can be challenging with limited data. Our work integrates user models with limited data with application-specific techniques to address some of these challenges. The dissertation describes steps for implementing accurate virtual stakeholder models based on limited training data. </p><p>Another method for dealing with limited data, based upon computing training data uncertainty, is also presented in this dissertation. Results presented show more stable convergence in fewer iterations when using an uncertainty-based incremental sampling method than when using stability based sampling or random sampling. The technique is described in additional detail. </p><p>The dissertation also discusses non-stationary reinforcement-based feature selection for the interactive optimization component of our system. The presented results indicate that the proposed feature selection approach can effectively mitigate against superfluous and adversarial dimensions which if left untreated can lead to degradation in both computational performance and interactive optimization performance against analytically determined environmental fitness functions. </p><p>The contribution of this dissertation lays the foundation for developing a framework for multi-stakeholder consensus decision-making in the presence of limited data.</p>

Concevoir une structure composite consiste à relever un défi de taille : alors qu'un ingénieur qui conçoit un produit mécanique à base de matériau métallique se concentre principalement sur le développement d'une forme qui garantira un comportement spécifique, l'ingénieur pour qui le problème de conception est celui d'un produit à base de matériaux composites doit trouver la meilleure combinaison forme - structure de matériau. Ainsi, il doit aussi concevoir simultanément un matériau et la topologie produit. La combinatoire s’avère être complexe et les espaces de solutions de très grande taille.Les outils de CAO et de simulation par éléments finis n'offrent pas au concepteur une approche permettant d'explorer les espaces de recherche de manière interactive et rapide. Le travail de thèse conduit à une nouvelle approche numérique permettant de manipuler chaque paramètre de conception caractérisant une structure composite, quelle que soit l’échelle à laquelle il est pertinent.Premièrement, le modèle de comportement paramétrique et réduit (Parametric and Reduced Behavior Model, PRBM) est un modèle dit séparé. Il permet :1- une approche multi-échelle : les paramètres mécaniques de la structure sont explicitement décrits comme issus de la qualité matérielle de chaque fibre, de la matrice, de chaque couche et de la topologie même du stratifié,2- une approche multi-physique: indépendamment le comportement mécanique de chaque couche et de chaque interface est traité pour donner lieu au comportement du stratifié. Des situations de comportements statiques et dynamiques sont étudiés. Dans le cas du comportement dynamique, le caractère visco-élastique est devenu un enjeu conceptuel.Deuxièmement, une méthode mixant dérivées non entières et usage de la méthode PGD a permis la réalisation du PRBM. Intégré dans un modèle de connaissance paramétrique (Parametric Knowledge Model, PKM) auprès de modèles de connaissances experts, il constitue la base d'une méthode interactive d’aide à la conception.Le PKM est traité par une méthode d'optimisation évolutionnaire. De ce fait, le concepteur peut explorer de façon interactive les espaces de conception. Pour qualifier nos modèles et notre PRBM, nous étudions 2 problèmes de conception de structures stratifiées. Les solutions déterminées sont qualifiées vis-à-vis de simulations par éléments finis ou selon une approche empirique. / The design process of laminated composites faces a major challenge: while an engineer designing a metallic based mechanical product is mainly focusing on the development of a shape that will guarantee a specific behavior, the engineer designing a composite based product must find the best combination of the shape-material structure. Therefore, he must simultaneously create a material and the product topology. The number of design solutions can be huge since the solution space is considerable.Standard CAE systems (CAD, Finite Element Simulation) do not provide an approach to explore these solution spaces efficiently and interactively. A new numerical procedure is proposed to allow engineers to handle each design parameter of a laminated composite structure, each at its relevant scale.First, the Parametric and Reduced Behavior Model (PRBM) is a separated model that enables reasoning based on1- A multiscale approach: the mechanical parameters of the structure are explicitly described as coming from the material quality of each fiber, the matrix, each layer and the topology of the laminate,2- A multiphysical approach: independently the mechanical behavior of each layer and each interface is processed, leading to the behavior of the laminate. Some situations of static and dynamic behavior are studied. In the case of dynamic behavior, the creeping becomes a conceptual issue.Secondly, a method mixing fractional derivatives and the Proper Generalized Decomposition (PGD) method allowed the creation of the PRBM. Integrated into a Parametric Knowledge Model (PKM) with other expert knowledge models, the PRBM makes the basis of an interactive method of design support.The PKM is processed by an evolutionary optimization method. As a result, the designer can interactively explore the design space. To qualify our models and our PRBM, we study two problems of design of laminated composite structures. The solutions determined are qualified versus finite element simulations or according to an empirical approach. / El diseño de una estructura compuesta es un desafío mayor, mientras que un ingeniero que diseña un producto mecánico con materiales metálicos se concentra principalmente en el desarrollo de una geometría que garantice un comportamiento específico, el ingeniero que diseña un producto con materiales compuestos debe encontrar la mejor combinación forma – estructura del material. De esta manera, el ingeniero debe diseñar simultáneamente el material y la topología del producto, razón por la que esta combinación se vislumbra compleja, puesto que los espacios de solución son gran tamaño.Las herramientas CAO y de simulación por elementos finitos no ofrecen al diseñador una metodología que permita explorar los espacios de solución de manera interactiva y rápida. Por lo tanto, este trabajo de tesis propone un nuevo enfoque numérico que permite manipular parámetros de diseño que caracterizan la estructura compuesta, cualquiera que sea la escala de pertinencia.Como primera medida, el modelo de comportamiento paramétrico y reducido (Parametric and Reduced Behavior Model, PRBM) es un modelo definido de manera separada que permite:1- Un enfoque multiescala: los parámetros mecánicos se presentan de manera explícita en términos de las propiedades de cada fibra, de la matriz, de cada capa y de la topología del mismo apilamiento.2- Un enfoque multifísico: el comportamiento mecánico de cada capa y cada interface se modela de manera independiente para dar lugar al comportamiento del apilamiento. Se estudian situaciones de casos de comportamiento estático y dinámico. En el caso dinámico en particular, se tiene en cuenta también la característica viscoelástica de las interfaces.Como segunda medida, un método que combina derivadas no enteras y el uso de la descomposición propia generalizada (PGD), permite la realización del PRBM. Este constituye la base de un método interactivo de ayuda al diseño, pues está integrado dentro de un modelo de conocimiento (PKM) que también incorpora mejores prácticas aprendidas por expertos.El PKM es utilizado por un método de optimización evolucionaria. De esta manera, el diseñador puede explorar de manera interactiva los espacios de solución. Para validar nuestros modelos y el PRBM, se estudian dos problemas de diseño de estructuras apiladas. Las soluciones obtenidas se comparan con respecto a simulaciones obtenidas por el método de los elementos finitos y con respecto a resultados experimentales.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2017-06-14T11:09:20Z No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Dissertação - Luiz Henrique Reis de Jesus - 2017.pdf: 3311759 bytes, checksum: 20e4044376c2666f102131892f7b3fc2 (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2017-06-14T11:09:35Z (GMT) No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Dissertação - Luiz Henrique Reis de Jesus - 2017.pdf: 3311759 bytes, checksum: 20e4044376c2666f102131892f7b3fc2 (MD5) / Made available in DSpace on 2017-06-14T11:09:35Z (GMT). No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Dissertação - Luiz Henrique Reis de Jesus - 2017.pdf: 3311759 bytes, checksum: 20e4044376c2666f102131892f7b3fc2 (MD5) Previous issue date: 2017-05-15 / Conselho Nacional de Pesquisa e Desenvolvimento Científico e Tecnológico - CNPq / This dissertation of master degree describes an approach of the interactive optimization process associate to the Evolution Strategies method for the evaluation of the loudspeaker optimization project, with the objective to present the advantages achieved after the user interventions throughout the optimization process. Its development is based on the methodology of the Evolution Strategies characterized by the concept of natural selection, which uses combination and mutation methods to generate new individuals. However, for greater efficiency in the responses of the objective function, as well as the reduction in its convergence time, the optimization process requires interventions in stabilization of local minima and maxima. In the interventions made available to the user, will be considered manipulations in the operators of mutation and combination, mutation of the parameters of self-adaptation, as well as the change of objective and the variation of their respective restriction. As a differential, an interface was developed to make feasible the user interventions applied to the optimization process, as well as the monitoring of the entire process. In this work, also evaluated optimization test functions with the objective of validating the proposed methodology. / Esta dissertação de mestrado descreve uma abordagem do processo de otimização interativa associado ao método das Estratégias de Evolução para a avaliação do projeto de otimização do alto-falante, com o objetivo de apresentar as vantagens alcançadas após as intervenções do usuário ao longo do processo de otimização. Seu desenvolvimento é baseado na metodologia das Estratégias de Evolução caracterizada pelo conceito de seleção natural, o qual utiliza de métodos de combinação e mutação para a geração de novos indivíduos. No entanto, para maior eficiência nas respostas a função objetivo, bem como a redução em seu tempo de convergência, o processo de otimização necessita de intervenções em estabilizações de mínimos e máximos locais. Nas intervenções disponibilizadas ao usuário, serão consideradas manipulações nos operadores de mutação e combinação, mutação dos parâmetros de auto-adaptação, bem como a mudança de objetivo e a variação de sua respectiva restrição. Como diferencial, foi desenvolvida uma interface para viabilizar as intervenções do usuário aplicadas ao processo de otimização, bem como o acompanhamento de todo o processo. Neste trabalho, também foram avaliadas funções de teste de otimização com o objetivo de validar a metodologia proposta.

本論文では人間の感性のモデルを関数と仮定し,最適化手法によってその感性の最適点を求めることで情報推薦を行うフレームワークについて研究を行った.論文中では,人間の感性に基づく関数の特性について明らかにし,人間の多様な感性への対応できる最適化アルゴリズムを開発した.また,感性に基づく探索に適した設計変数空間の半自動生成手法を開発し,脳機能情報を用いた感性のモデル化について検討した. / 博士(工学) / Doctor of Philosophy in Engineering / 同志社大学 / Doshisha University

Indiana University-Purdue University Indianapolis (IUPUI) / Environmental planning is very cumbersome work for environmentalists, government agencies like USDA and NRCS, and farmers. There are a number of conflicts and issues involved in such a decision making process. This research is based on the work to provide a common platform for environmental planning called WRESTORE (Watershed Restoration using Spatio-Temporal Optimization of Resources). We have designed a system that can be used to provide the best management practices for environmental planning. A distributed system was designed to combine high performance computing power of clusters/supercomputers in running various environmental model simulations. The system is designed to be a multi-user system just like a multi-user operating system. A number of stakeholders can log-on and run environmental model simulations simultaneously, seamlessly collaborate, and make collective judgments by visualizing their landscapes. In the research, we identified challenges in running such a system and proposed various solutions. One challenge was the lack of fast optimization algorithm. In our research, several algorithms are utilized such as Genetic Algorithm (GA) and Learning Automaton (LA). However, the criticism is that LA has a slow rate of convergence and that both LA and GA have the problem of getting stuck in local optima. We tried to solve the multi-objective problems using LA in batch mode to make the learning faster and accurate. The problems where the evaluation of the fitness functions for optimization is a bottleneck, like running environmental model simulation, evaluation of a number of such models in parallel can give considerable speed-up. In the multi-objective LA, different weight pair solutions were evaluated independently. We created their parallel versions to make them practically faster in computation. Additionally, we extended the parallelism concept with the batch mode learning. Another challenge we faced was in User Modeling. There are a number of User Modeling techniques available. Selection of the best user modeling technique is a hard problem. In this research, we modeled user's preferences and search criteria using an ANN (Artificial Neural Network). Training an ANN with limited data is not always feasible. There are many situations where a simple modeling technique works better if the learning data set is small. We formulated ways to fine tune the ANN in case of limited data and also introduced the concept of Deep Learning in User Modeling for environmental planning system.