Thesis (M.H.A.)--University of Michigan, 1962.

No description available.

Cette thèse sur travaux composée de quatre articles s’intéresse aux chaînes logistiques à deux échelons comportant un fournisseur en situation de monopole et de N clients dont les demandes pour un même produit sont corrélées. Les trois premiers articles étudient l’impact simultané des coopérations verticale et horizontale sur la performance globale. L’article 1 propose des extensions du modèle d’optimisation des stocks de Zhu et Thonemann (2004) en complétant notamment la coopération verticale entre le fournisseur et ses clients par une alliance horizontale et un échange d’informations entre clients. L’article 2 propose une modélisation multi-agents individu-centrée afin d’étudier l’impact sur la performance et la stabilité de la chaîne, de la diversité des comportements des clients face au risque et de leurs règles d’interaction dans le cadre d’une possible distorsion de l’information échangée. Nous proposons ensuite deux articles à visée plus applicative. L’article 3 s’intéresse à l’industrie pharmaceutique et étudie l’influence de la coopération entre grossistes-répartiteurs socialement responsables et leur laboratoire fournisseur pour réduire les surstocks et les gaspillages. L’article 4 consiste à comparer un pilotage centralisé à un pilotage décentralisé des stocks suite à un choc important de demande d’un produit alimentaire périssable. Les résultats de simulation font ressortir des conditions d’équilibre ainsi que des recommandations sur le pilotage global de ce type de chaînes. Plus généralement, cette thèse a permis de montrer l’intérêt d’une approche connexionniste de chaînes logistiques complexes avec agents hétérogènes s’échangeant de l’information. / AThis thesis is composed by two papers focusing on two-level supply chains with a monopoly supplier and N clients whose demands for the same products are correlated. The first paper studies the simultaneous impact of vertical and horizontal cooperation on the overall performance and the stability of the chain. It proposes an individual-centered multi-agents approach for studying the impact on performance of different kinds of customer behaviors associated with overstock risk and their interaction rules under possible distortion of the exchanged information. The second paper deals with pharmaceutical distribution and studies the influence of the cooperation between wholesale distributors and a unique supplier to reduce overstocks and wastes. More generally, this thesis has shown the relevance of a connectionist approach of complex supply chains with heterogeneous agents exchanging information.

The problem of blind separation refers to recovering original signals, called source signals, from the mixed signals, called observation signals, in a reverberant environment. The mixture is a function of a sequence of original speech signals mixed in a reverberant room. The objective is to separate mixed signals to obtain the original signals without degradation and without prior information of the features of the sources. The strategy used to achieve this objective is to use multiple bands that work at a lower rate, have less computational cost and a quicker convergence than the conventional scheme. Our motivation is the competitive results of unequal-passbands scheme applications, in terms of the convergence speed. The objective of this research is to improve unequal-passbands schemes by improving the speed of convergence and reducing the computational cost. The first proposed work is a novel maximally decimated unequal-passbands scheme. This scheme uses multiple bands that make it work at a reduced sampling rate, and low computational cost. An adaptation approach is derived with an adaptation step that improved the convergence speed. The performance of the proposed scheme was measured in different ways. First, the mean square errors of various bands are measured and the results are compared to a maximally decimated equal-passbands scheme, which is currently the best performing method. The results show that the proposed scheme has a faster convergence rate than the maximally decimated equal-passbands scheme. Second, when the scheme is tested for white and coloured inputs using a low number of bands, it does not yield good results; but when the number of bands is increased, the speed of convergence is enhanced. Third, the scheme is tested for quick changes. It is shown that the performance of the proposed scheme is similar to that of the equal-passbands scheme. Fourth, the scheme is also tested in a stationary state. The experimental results confirm the theoretical work. For more challenging scenarios, an unequal-passbands scheme with over-sampled decimation is proposed; the greater number of bands, the more efficient the separation. The results are compared to the currently best performing method. Second, an experimental comparison is made between the proposed multiband scheme and the conventional scheme. The results show that the convergence speed and the signal-to-interference ratio of the proposed scheme are higher than that of the conventional scheme, and the computation cost is lower than that of the conventional scheme.

Made available in DSpace on 2016-08-29T15:32:30Z (GMT). No. of bitstreams: 1 tese_2759_DissertacaoMestradoBrunoZanettiMelotti.pdf: 1283815 bytes, checksum: 1948186d38f6639b57cb0387d5eaba2c (MD5) Previous issue date: 2009-11-27 / Dado um documento para categorização, um sistema de categorização multi-rótulo de texto tipicamente ordena um conjunto de categorias pré-definido, de acordo com a adequação delas ao documento, e seleciona as categorias do topo do ranking como o conjunto de categorias do documento. Empates no ranking eventualmente existentes podem ser tratados de diferentes maneiras, mas, muito embora isso possa afetar as métricas utilizadas para avaliar o desempenho dos categorizadores multi-rótulo de texto, este problema parece ter sido pouco estudado na literatura. Neste trabalho, analisamos o impacto de diferentes tipos de ranking sobre diversas métricas de avaliação de desempenho de categorizadores multi-rótulo de texto, a saber: one-error, coverage, ranking loss, average precision, R-precision, Hamming loss, exact match, precision, recall, e 1 F . Para isso, reformulamos sua definição de modo a considerar empates de acordo com o tipo de ranking empregado. Utilizamo-las então para avaliar o desempenho das técnicas de categorização multi-rótulo de texto k -vizinhos mais próximos ( k NN), k -vizinhos mais próximos multi-rótulo (ML- k NN), rede neural sem peso do tipo VG-RAM (VG-RAM WNN) e VG-RAM com correlação de dados (VG-RAM WNNCOR) na categorização de duas bases multi-rótulo de texto com grande número de categorias (105 e 692 categorias). Descobrimos que, dependendo do tipo de ranking empregado, os resultados de desempenho são significativamente diferentes para muitas das métricas analisadas, o que sugere que o tipo de ranking deve ser claramente indicado na avaliação de técnicas de categorização multi-rótulo de texto.

abstract: In many fields one needs to build predictive models for a set of related machine learning tasks, such as information retrieval, computer vision and biomedical informatics. Traditionally these tasks are treated independently and the inference is done separately for each task, which ignores important connections among the tasks. Multi-task learning aims at simultaneously building models for all tasks in order to improve the generalization performance, leveraging inherent relatedness of these tasks. In this thesis, I firstly propose a clustered multi-task learning (CMTL) formulation, which simultaneously learns task models and performs task clustering. I provide theoretical analysis to establish the equivalence between the CMTL formulation and the alternating structure optimization, which learns a shared low-dimensional hypothesis space for different tasks. Then I present two real-world biomedical informatics applications which can benefit from multi-task learning. In the first application, I study the disease progression problem and present multi-task learning formulations for disease progression. In the formulations, the prediction at each point is a regression task and multiple tasks at different time points are learned simultaneously, leveraging the temporal smoothness among the tasks. The proposed formulations have been tested extensively on predicting the progression of the Alzheimer's disease, and experimental results demonstrate the effectiveness of the proposed models. In the second application, I present a novel data-driven framework for densifying the electronic medical records (EMR) to overcome the sparsity problem in predictive modeling using EMR. The densification of each patient is a learning task, and the proposed algorithm simultaneously densify all patients. As such, the densification of one patient leverages useful information from other patients. / Dissertation/Thesis / Ph.D. Computer Science 2014

Nas metodologias de mapeamento de QTLs tradicionais, a relação de causalidade entre os caracteres fenotípicos e QTLs normalmente não são consideradas. O desenvolvimento deste trabalho contou com a utilização de dados longitudinais de crescimento de progênies oriundas do cruzamento entre os parentais PB217 e PR255 de um plantio de seringueira, localizado em uma área com dois períodos bem definidos ao longo do ano (altas e médias temperaturas; altas e baixas taxas precipitação). O experimento contém 4 medidas de incremento em diâmetro e altura, que são componentes periódicos do crescimento total da cultura, mensurados em um intervalo de dois anos (entre os 18 aos 52 meses de idade das plantas), sendo dois períodos em estação climática favorável ao desenvolvimento e dois em estação desfavorável, intercalados. Dessa forma foram estudados os parâmetros de relacionamento fenotípico e genético com objetivo de construir um diagrama de arquitetura genética que pondere relações de causalidade. Para modelar os dados fenotípicos foi realizado um elaborado modelo multi-caracteres que contemplou a variação espacial das parcelas experimentais e a variação entre os períodos de medição. Para tanto, foram ajustadas matrizes de variância-covariância (VCOV) adequadas à realidade dos dados, e incorporados dados meteorológicos que descrevessem cada um dos períodos. A partir destes modelos, os valores genotípicos ajustados foram utilizados na detecção dos QTLs. Posteriormente, fenótipos e genótipos foram articulados em um diagrama causal estrutural capaz de inferir sobre padrões genéticos de comportamento de crescimento da cultura. Foram mapeados um total de 13 QTLs, sendo que dois deles foram coincidentes para componentes periódicos de diâmetro nos períodos de estação desfavorável. Foi possível identificar efeitos aditivos e devido à dominância interessantes para o desenvolvimento em períodos de menores temperaturas, apontar o parental PR255 como portador de alelos importantes no desenvolvimento em clima adverso, estimar efeitos indiretos de QTLs não mapeados para determinadas características e explicar o padrão comportamental de crescimento no período em que as progênies foram avaliadas. Esta abordagem demonstrou-se proficiente para utilização em programas melhorando genético assistido por marcadores, por agregar informações pertinentes à seleção dos melhores materiais genéticos. / In traditional methodologies of QTL mapping, the causal relationship between phenotypic characters and QTLs are usually not considered. The development of this work involved the use of longitudinal growth data of progenies from parental PR255 and PB217 of a rubber tree plantation, located in an area with two periods of high and medium temperature and low and high precipitation rates well defined throughout the year. The experiment contains four measures of increment of diameter and height, which are periodic growth components of the total crop growing at an interval of two years (from 18 to 52 months old plants), two periods in a favorable climate station and two in a adverse station, intercalated. Was studied the parameters of phenotypic and genetic relationships in order to construct a diagram of genetic architecture to examine these causal relationships. A multi-trait-multi-occasion model that take into consideration spatial variation and climatic variation was developed. It also contains a variation-covariation matrix with appropriate to the reality of the data were adjusted and incorporated meteorological data was conducted to describe each of the periods. From these models the adjusted genotypic values were used in the detection of QTLs and later phenotypes and genotypes were linked in a structural causal diagram to infer about the genetic patterns of behavior. A total of 13 QTLs were mapped to the periodic growth components and total growth. The genetic architecture was able to identify additive effects and effects due dominance interesting to the development in periods of lower temperatures and drought, pointing parental PR255 as carrier of important alleles to development in adverse weather, estimating indirect effects of QTLs that were not mapped to certain characteristics and explain the physiological behavior pattern of growth in the period in which progenies were evaluated. This approach proved to be proficient to use in breeding programs aiming to implement marker assisted selection.

M.Sc. / Advancements in technology used in financial markets have led to substantial automation of tasks within the financial industry. Data analysis, trade execution and trade processing have been automated, reducing costs and increasing productivity. Algorithmic trading is the automated execution of trades on an electronic trading platform; it has been used to gain competitive advantage in financial markets since the early 1990s. Algorithmic trading applications, which must analyse information and determine whether to buy or sell, are well suited to the use of autonomous software agents. Multi-agent systems are better suited to the increasing complexity of algorithmic trading systems and the flexibility required by rapidly changing markets than single-agent systems. The granularity of components (agents) in multi-agent systems also promotes reuse and simplifies individual agent design. Algorithmic trading is, however, subject to challenges specifically in terms of data volume, speed of access and speed of processing. In order to utilise a multi-agent system solution the interactions between agents which allow distributed problem solving must be as efficient as possible. This dissertation investigates the use of indirect coordination to improve the efficiency of interactions between agents in multi-agent systems and to simplify agent design. Indirect coordination utilises environment abstractions known as artefacts to facilitate interaction between agents; such interaction can be simple data transfer or requests, complex coordination protocols as well as negotiation protocols. The investigation resulted in a framework that allows agents to transition between direct and indirect interaction techniques based on the specific interaction task at hand. The framework is built on two existing platforms, ii Java Agent DEvelopment Framework (JADE) and Common ARTifact Infrastructure for AGents Open environments (CARTAGO). These platforms are combined into the JADE-CARTAGO Algorithmic Trading (JCAT) framework that provides the infrastructure needed for both direct and indirect interactions. Investigations into the performance of the JCAT framework have shown that artefacts improve interaction efficiency by reducing data loss in tasks such as information publishing, and perform as well as direct communication within certain constraints for other tasks. When limiting the number of agents in an interaction to 50 agents, artefacts perform at least as well as direct communication using agent communication language messages.

La rivière Marque est un petit cours d’eau localisé sur un bassin versant péri-urbain du Nord de la France, traversant des zones agricoles et urbaines. Au niveau de Villeneuve d’Ascq, cette rivière reçoit les eaux de deux principaux ouvrages de traitement de l’eau : (i) les eaux urbaines traitées de la station d’épuration de Villeneuve d’Ascq ; et (ii) les eaux du lac du Héron qui est un bassin d’orage. Afin d’améliorer la compréhension de la rivière dans cette zone, des suivis basse et haute fréquences ont été mis en place en 2014. En complément, une bouée instrumentée appartenant à l’Agence de l’Eau Artois-Picardie a été déployée en 2015 sur le lac. Une approche multi-traceurs a été envisagée pour tenter d’identifier les différentes pressions sur La Marque avec l’étude des paramètres physico-chimiques (O2, pH, conductivité, MES, COD), des nutriments (NO3-, NH4+, PO43-), des micropolluants métalliques (Cu, Pb, Zn) et organiques (CAF, CBZ) et des conditions physiques (débit, pluviométrie). Les points clés mis en évidence sont les suivants : (1) La Marque est fortement impactée par le rejet d’eaux usées non traitées dû à des défaillances dans les réseaux d’assainissement ; (2) la station d’épuration contribue significativement à enrichir La Marque en CBZ et en Zn, ce qui participe vis-à-vis de ce dernier à la dégradation de l’état écologique ; (3) la capacité de traitement par décantation et processus biogéochimiques naturels du lac du Héron est efficace, cependant il est nécessaire d’entreprendre un curage du chenal d’entrée ; et (4) le rejet du lac participe à améliorer ponctuellement mais faiblement la qualité de la rivière, excepté lors d’évènements particuliers. / The Marque River is a little watercourse located in a peri-urban watershed in Northern France, going through agricultural and urban areas. Nearby Villeneuve d’Ascq, this river receives waters from two main water treatment plants: (i) the waste water treatment plant of Villeneuve d’Ascq; and (ii) the waters from a stormwater basin, the Heron lake. In order to improve the understanding of the functioning of the river in this area, low and high frequencies monitoring have been undertaken in 2014. In addition, an instrumented buoy belonging to the French Water Agency Artois-Picardie was implemented in 2015 on the lake. A multi-tracers approach was considered in order to identify different pressures on The Marque River including physicochemical parameters (O2, pH, conductivity, SPM, DOC), nutrients (NO3-, NH4+, PO43-), trace metals (Cu, Pb, Zn), organic micro-pollutants (CAF, CBZ) and hydrodynamic conditions (water flow, rainfall). The key points that have been highlighted are as follows: (i) The Marque River is strongly impacted by the discharge of wastewaters, due to sanitation network failures; (ii) the waste water treatment plant contributes significantly to the enrichment of The Marque River in CBZ and Zn, the latter participating to the degradation of the ecological status; (iii) the treatment capacity by naturals settling and biogeochemical processes is efficient within the Heron lake. However, the dredging of the entrance channel must be undertaken; and (iv) the discharge from the lake to the river tends to ameliorate scarcely and punctually the water quality of the river, excepted during special events.

Many things can be communicated through dialogue, including information, thoughts, feelings, attitudes, beliefs and personal experiences. More recently, dialogues have been used in focus group research and in program evaluations. Despite the increasing prevalence of dialogue in research and evaluation, much is still unknown about dialogue, especially how dialogue emerges and occurs within a group setting. The aim of the current study was to describe and identify the various factors involved in a dialogue, examine the relationships among these factors, and conceptualize the process of dialogue within a multi-stakeholder participatory evaluation. A qualitative analysis of three focus groups, each comprised of eight to ten participants, yielded several findings. First, several factors were found to help facilitate the interactions between multiple stakeholders in dialogue, including the development of common ground and specific contributions made by participants. Secondly, communication within these multiple stakeholder groups was found to alternate between two individuals, a dyadic exchange, or between multiple participants, a complex exchange. Thirdly, the moderator and participants were found to take on each other roles. Finally, from these conversations, a model was developed to illustrate the progression of a dialogue in these groups. These results have many implications for program evaluators, focus group leaders, and other practitioners in the field.