<p> Object understanding is a fundamental problem in computer vision and it has been extensively researched in recent years thanks to the availability of powerful GPUs and labelled data, especially in the context of images. However, 3D object understanding is still not on par with its 2D domain and deep learning for 3D has not been fully explored yet. In this dissertation, I work on two approaches, both of which advances the state-of-the-art results in 3D classification and segmentation.</p><p> The first approach, called MVRNN, is based multi-view paradigm. In contrast to MVCNN which does not generate consistent result across different views, by treating the multi-view images as a temporal sequence, our MVRNN correlates the features and generates coherent segmentation across different views. MVRNN demonstrated state-of-the-art performance on the Princeton Segmentation Benchmark dataset.</p><p> The second approach, called PointGrid, is a hybrid method which combines points and regular grid structure. 3D points can retain fine details but irregular, which is challenge for deep learning methods. Volumetric grid is simple and has regular structure, but does not scale well with data resolution. Our PointGrid, which is simple, allows the fine details to be consumed by normal convolutions under a coarser resolution grid. PointGrid achieved state-of-the-art performance on ModelNet40 and ShapeNet datasets in 3D classification and object part segmentation. </p><p>

<p> There have been a number of studies that try to predict sporting event outcomes. Most previous research has involved results in football and college basketball. Recent years has seen similar approaches carried out in professional basketball. This thesis attempts to build upon existing statistical techniques and apply them to the National Basketball Association using a synthesis of algorithms as motivation. A number of ensemble learning methods will be utilized and compared in hopes of improving the accuracy of single models. Individual models used in this thesis will be derived from Logistic Regression, Na&iuml;ve Bayes, Random Forests, Support Vector Machines, and Artificial Neural Networks while aggregation techniques include Bagging, Boosting, and Stacking. Data from previous seasons and games from both?players and teams will be used to train models in R.</p><p>

Made available in DSpace on 2019-04-05T23:09:11Z (GMT). No. of bitstreams: 0 Previous issue date: 2007-09-17 / Preference information filtering is a required feature of multi-agent systems if they need accessing a database on Semantic Web. However, few studies have paid attention to creating a generic representation of preference for agent consumption. In this work, we describe our proposal for representing preferences in a declarative, domain-independent and machine-interpretable way in OWL. The aim of OWLPREF is to be used for knowledge sharing among agents, allowing the representation of partial order between classes, attributes and attribute-values. We also describe the API that translates OWLPREF into SPARQL queries, as well as exemplify the use of the ontology and of the API in the development of multi-agent application in the Semantic Web context to demonstrate OWLPREF capabilities of sharing and representation of preferences. / Filtrar informações baseadas em preferências é uma característica essencial de sistemas multiagentes se estes necessitam acessar dados na Web Semântica. Entretanto, poucos trabalhos têm estudado a questão da representação genérica de preferências para uso por esses agentes. O objetivo deste trabalho é descrever nossa proposta de representação de preferências de maneira declarativa, independente de domínio e interpretável por máquinas usando OWL. OWLPREF será útil para o compartilhamento de conhecimento entre agentes de software, permitindo a representação parcial de ordem entre atributos, classes e valores de atributos. Neste trabalho descrevemos a interface de software que mapeia OWLPREF em consultas SPARQL e a exemplificamos em uma aplicação na Web Semântica, demonstrando sua capacidade de representação e compartilhamento de preferências.

Nowadays, millions of pictures are shared through the internet without applying any authentication system. This may cause serious problems, particularly in situations where the digital image is an important component of the decision making process for example, child pornography and movie piracy. Motivated by this, the present research investigates the performance of estimating Photo Response Non-Uniformity (PRNU) and developing new estimation approaches to improve the performance of digital source camera identification. The PRNU noise is a sensor pattern noise characterizing the imaging device. Nonetheless, the PRNU estimation procedure is faced with the presence of image-dependent information as well as other non-unique noise components. This thesis primarily focuses on efficiently estimating the physical PRNU components during different stages. First, an image sharpening technique is proposed as a pre-processing approach for source camera identification. The sharpening method aims to amplify the PRNU components for better estimation. In the estimation stage, a new weighted averaging (WA) technique is presented. Most existing PRNU techniques estimate PRNU using the constant averaging of residue signals extracted from a set of images. However, treating all residue signals equally through constant averaging is optimal only if they carry undesirable noise of the same variance. Moreover, an improved version of the locally adaptive discrete cosine transform (LADCT) filter is proposed in the filtering stage to reduce the effect of scene details on noise residues. Finally, the post-estimation stage consists of combining the PRNU estimated from each colour plane aims to reduce the effect of colour interpolation and increasing the amount of physical PRNU components. The aforementioned techniques have been assessed on two image datasets acquired by several camera devices. Experimental results have shown a significant improvement obtained with the proposed enhancements over related state-of-the-art systems. Nevertheless, in this thesis the experiments are not including images taken with various acquisition different resolutions to evaluate the effect of these settings on PRNU performance. Moreover, images captured by scanners, cell phones can be included for a more comprehensive work. Another limitation is that investigating how the improvement may change with JPEG compression or gamma correction. Additionally, the proposed methods have not been considered in cases of geometrical processing, for instance cropping or resizing.

<p> Being able to accomplish tasks with multiple learners through learning has long been a goal of the multiagent systems and machine learning communities. One of the main approaches people have taken is reinforcement learning, but due to certain conditions and restrictions, applying reinforcement learning in a multiagent setting has not achieved the same level of success when compared to its single agent counterparts. </p><p> This thesis aims to make coordination better for agents in cooperative games by improving on reinforcement learning algorithms in several ways. I begin by examining certain pathologies that can lead to the failure of reinforcement learning in cooperative games, and in particular the pathology of <i> relative overgeneralization</i>. In relative overgeneralization, agents do not learn to optimally collaborate because during the learning process each agent instead converges to behaviors which are robust in conjunction with the other agent's exploratory (and thus random), rather than optimal, choices. One solution to this is so-called <i>lenient learning</i>, where agents are forgiving of the poor choices of their teammates early in the learning cycle. In the first part of the thesis, I develop a lenient learning method to deal with relative overgeneralization in independent learner settings with small stochastic games and discrete actions. </p><p> I then examine certain issues in a more complex multiagent domain involving parameterized action Markov decision processes, motivated by the RoboCup 2D simulation league. I propose two methods, one batch method and one actor-critic method, based on state of the art reinforcement learning algorithms, and show experimentally that the proposed algorithms can train the agents in a significantly more sample-efficient way than more common methods. </p><p> I then broaden the parameterized-action scenario to consider both repeated and stochastic games with continuous actions. I show how relative overgeneralization prevents the multiagent actor-critic model from learning optimal behaviors and demonstrate how to use Soft Q-Learning to solve this problem in repeated games. </p><p> Finally, I extend imitation learning to the multiagent setting to solve related issues in stochastic games, and prove that given the demonstration from an expert, multiagent Imitation Learning is exactly the multiagent actor-critic model in Maximum Entropy Reinforcement Learning framework. I further show that when demonstration samples meet certain conditions the relative overgeneralization problem can be avoided during the learning process.</p><p>

A widely adopted approach to solving constraint satisfaction problems combines backtracking tree search with various degrees of constraint propagation for pruning the search space. One common technique to improve the execution efficiency is to add redundant constraints, which are constraints logically implied by others in the problem model and may offer extra information to enhance constraint propagation. However, some redundant constraints are propagation redundant and hence do not contribute additional propagation information to the constraint solver. In this thesis, we propose propagation rules as a tool to compare the propagation strength of constraints, and establish results relating logical and propagation redundancy. / Redundant constraints arise naturally in the process of redundant modeling, where two models of the same problem are connected and combined through channeling constraints. We characterize channeling constraints in terms of restrictive and unrestrictive channel function and give general theorems for proving propagation redundancy of constraints in the combined model. We illustrate, on problems from CSPLib, how detecting and removing propagation redundant constraints can often significantly speed up constraint-solving. / Choi Chiu Wo. / "September 2005." / Adviser: Jimmy Lee. / Source: Dissertation Abstracts International, Volume: 67-07, Section: B, page: 3890. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2005. / Includes bibliographical references (p. 106-117). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract in English and Chinese. / School code: 1307.

INTRODUÇÃO: A ressussitação volêmica é uma terapia frequentemente usada em pacientes criticamente enfermos com falência circulatória aguda. O benefício hemodinâmico esperado é um aumento no volume de ejeção do ventrículo esquerdo e, portanto, no débito cardíaco. Em pacientes com falência circulatória aguda, a taxa média de respondedores à expansão volêmica é de aproximadamente 50%. São usados à beira do leito parâmetros estáticos e dinâmicos. Os parâmetros dinâmicos apresentam melhor correlação com a resposta a desafio hídrico do que os estáticos. Em particular, o ΔPP é o mais acurado em pacientes ventilados com volume corrente (VAC) ≥ 8ml/kg do peso ideal, mas não com volumes correntes menores. Portanto, este estudo foi desenhado para avaliar a variação da pressão de pulso (ΔPP) como preditor de responsividade a volume em pacientes ventilados com volumes correntes reduzidos (< 8ml/kg do peso ideal). MÉTODOS: Estudo transversal, não-intervencionista, realizado no Hospital de Clínicas de Porto Alegre. Foram incluídos 38 pacientes internados no CTI adulto, de ambos os sexos, com idade ≥ 16 anos, sedados e em ventilação mecânica invasiva, monitorizados com linha arterial e catéter de artéria pulmonar, que necessitariam receber volume. Um paciente foi excluído por apresentar parada cardiorrespiratória durante a execução do protocolo. Foram feitas medidas hemodinâmicas (ΔPP, pressão arterial média [PAM], pressão venosa central [PVC], pressão média da artéria pulmonar [PMAP], pressão de oclusão da artéria pulmonar [POAP], débito cardíaco [DC]) e ventilatórias (volume corrente expiratório [VACexp], pressão de platô [Ppl], pressão de pico [Ppico], complacência estática [Cest], driving pressure [DP], pressão expiratória final total [PEEPtot]) antes e após o desafio hídrico. RESULTADOS: Dos trinta e sete pacientes, 17 apresentaram um aumento do índice cardíaco ≥ 15% após o desafio hídrico (respondedores) e 20 pacientes apresentaram um aumento do índice cardíaco < 15% (não respondedores). Todos os pacientes, com exceção de um, que apresentaram ΔPP ≥ 10% foram respondedores. O melhor ponto de corte encontrado foi 10% (área sob a curva ROC 0,74, sensibilidade 53%, especificidade 95%, Likelihood ratio positivo 9,3 e negativo 0,34). Corrigindo o ΔPP pela driving pressure (DP), o resultado foi semelhante, com área sob a curva ROC 0,76. Dos 37 pacientes incluídos, 25 estavam em choque séptico. Para o ponto de corte 10%, a área sob a curva ROC encontrada foi 0,84, com sensibilidade 77,8% e especificidade 93,3%. CONCLUSÃO: De acordo com os resultados obtidos nesse estudo, a variação da pressão de pulso teve valor limitado como preditor de resposta a volume em pacientes ventilados com volumes correntes reduzidos. O melhor ponto de corte encontrado foi 10%. Embora um ΔPP baixo não contraindique o desafio hídrico, um ΔPP ≥10% pode auxiliar na identificação de respondedores com choque séptico.

A grande maioria dos pacientes internados nas unidades de terapia intensiva necessita de suporte ventilatório durante sua internação. Estes pacientes acabam sendo expostos a eventos deletérios, como a alteração do mecanismo muco-ciliar, e, dessa forma, ao acumulo de secreções brônquicas. Uma maneira de resolver ou ao menos amenizar estas alterações é a aplicação de manobras de higiene brônquica, por exemplo: a vibrocompressão, que visa o deslocamento da secreção brônquica e sua depuração. Nem todos os pacientes mecanicamente ventilados respondem bem a essa manobra, o que leva os profissionais utilizarem técnicas de hiperinsuflação dos pulmões. Dentre as hiperinsuflações destaca-se a alteração no modo ventilatório pressão suporte. Visando comparar estas duas técnicas de higiene brônquica quanto a retirada de secreção em pacientes em ventilação mecânica (VM): vibrocompressão (VB) x vibrocompressão associada ao acréscimo de 10 cm H2O na pressão suporte (VB+PSV). Foi realizado um ensaio clínico randomizado incluindo pacientes em VM >48 horas internados no Centro de Terapia Intensiva do Hospital de Clínicas de Porto Alegre. Os pacientes foram alocados em dois grupos VB (G1) e VB + PSV(G2). Foram medidos os parâmetros hemodinâmicos e pulmonares e a quantidade de secreção aspirada (SEC) em dois momentos: 1. Aspiração isolada (ASP), 2. Associada a VB (G1) ou VB + PSV (G2), comparando-se durante estes momentos e entre os grupos, a variação desses parâmetros bem como a SEC. Durante este estudo foram incluídos 66 pacientes, 32 no G1 e 34 no G2. As características clínicas em ambos os grupos foram semelhantes. A SEC média aumentou de 1,3 ± 1,2 gramas (g) para 1,7 ± 1,6 g no G1 (p=0,102) e de 2,6 ± 3,0 g para 3,5 ± 3,8 g no G2 (p=0,018), a variação do volume corrente expirado (ΔVCE) e a variação da complacência dinâmica (ΔCdin) aumentaram em média no G2 de 16,2 ± 69,3 ml para 55,6 ± 69,2 ml, p=0,018, e de 0,01 ± 4,9 cm H2O para 2,8 ± 4,5 cm H2O, p=0,005, respectivamente. Tanto a VB quanto a VB + PSV demonstraram aumento da SEC comparados com a ASP, mas somente pós VB + PSV esse aumento foi significativo; além disso o ΔVCE e o ΔCdin aumentaram significativamente pós VB + PSV.

Wu, Yi. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2011. / Includes bibliographical references (p. 74-80). / Abstracts in English and Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Constraint Satisfaction Problems --- p.1 / Chapter 1.2 --- Weighted Constraint Satisfaction Problems --- p.3 / Chapter 1.3 --- Motivation and Goal --- p.4 / Chapter 1.4 --- Outline of the Thesis --- p.5 / Chapter 2 --- Background --- p.7 / Chapter 2.1 --- Constraint Satisfaction Problems --- p.7 / Chapter 2.1.1 --- Backtracking Tree search --- p.8 / Chapter 2.1.2 --- Local consistencies in CSP --- p.11 / Chapter 2.2 --- Weighted Constraint Satisfaction Problems --- p.18 / Chapter 2.2.1 --- Branch and Bound Search --- p.20 / Chapter 2.2.2 --- Local Consistencies in WCSP --- p.21 / Chapter 2.3 --- Global Constraints --- p.31 / Chapter 3 --- Tractable Projection-Safety --- p.36 / Chapter 3.1 --- Tractable Projection-Safety: Definition and Analysis --- p.37 / Chapter 3.2 --- Polynomially Decomposable Soft Constraints --- p.42 / Chapter 4 --- Examples of Polynomially Decomposable Soft Global Constraints --- p.48 / Chapter 4.1 --- Soft Among Constraint --- p.49 / Chapter 4.2 --- Soft Regular Constraint --- p.51 / Chapter 4.3 --- Soft Grammar Constraint --- p.54 / Chapter 4.4 --- Max_Weight/Min Weight Constraint --- p.57 / Chapter 5 --- Experiments --- p.61 / Chapter 5.1 --- The car Sequencing Problem --- p.61 / Chapter 5.2 --- The nonogram problem --- p.62 / Chapter 5.3 --- Well-Formed Parenthesis --- p.64 / Chapter 5.4 --- Minimum Energy Broadcasting Problem --- p.64 / Chapter 6 --- Related Work --- p.67 / Chapter 6.1 --- WCSP Consistencies --- p.67 / Chapter 6.2 --- Global Constraints . --- p.68 / Chapter 7 --- Conclusion --- p.71 / Chapter 7.1 --- Contributions --- p.71 / Chapter 7.2 --- Future Work --- p.72 / Bibliography --- p.74

Esta tese propõe um modelo probabilístico geral para a localização cooperativa de multirrobôs. O problema da localização multirrobôs pode ser definido como: dado um modelo do ambiente, estimar a localização de cada robô em um grupo atuando em um mesmo ambiente, com base nas informações sensoriais oferecidas pelas medidas de odometria, medidas do ambiente e detecções. Detecção é a habilidade de um robô identificar outro e determinar a distância relativa entre eles. A idéia principal da localização cooperativa de multirrobôs consiste em integrar medidas coletadas por diferentes robôs, de modo que todos possam se beneficiar dos dados adquiridos pelos outros robôs do grupo. Desta forma, detecções podem ser usadas para refinar a estimativa de postura de cada robô com base nas estimativas dos outros. Comunicação fornece aos robôs a habilidade de compartilhar suas crenças de postura de forma que possam cooperar para melhorar a acurácia da localização. Aqui é explorado o uso de diferentes tipos de informação para comunicar entre os robôs: propagação da detecção positiva, detecção negativa e multidetecção, os quais são integrados em um novo algoritmo, chamado Localização de Markov para Multirrobôs Cooperativos (LMMC). Também é proposto um protocolo de comunicação para a troca de dados entre os robôs e um conjunto de critérios que possibilitam a redução da comunicação por meio da diminuição da quantidade de dados trocados entre robôs, de um modo eficaz e eficiente. Os experimentos realizados em ambientes simulados demonstram que a abordagem proposta pode conduzir a resultados significativamente melhores de localização quando comparada à abordagem com detecção única e ainda com uma menor quantidade de mensagens trocadas entre os robôs. / In this thesis we propose a general probabilistic model to cooperative multirobot localization. The multirobot localization problem can be stated as follows: given a model of the environment, estimate the location of each robot in a group within the same environment based on sensors information that provides odometric measurements, environment measurements, and detections. Detection is the ability of one robot to identify others and to determine the relative location of other robots relative to its own. The key idea behind the multirobot localization approach is to integrate measurements taken by different robots, so that each one can benefit from data gathered by other robots in a group. In this sense, detections can be used to refine the pose estimates of a robot based on the other\'s estimate. Communication provides robots with the ability to exchange their pose beliefs, so that they can cooperate in order to improve their localization accuracy. We explore the use of different types of information to exchange among robots: propagation of positive detection, negative detection, and multidetection, which are integrated in a new algorithm, called Cooperative Multirobot Markov Localization, CMML. We also contribute a communication protocol that deals with the data transmitted among robots, and a set of communication rules that aims at reducing the amount of data exchanged among robots in an effective and efficient way. Experimental results, carried out in simulated environments, demonstrate that our approach can yield better localization results than a single-detection approach, at significantly smaller communication overhead.