Dissertations / Theses on the topic 'Style GAN'

Background: Generative Adversarial Networks (Goodfellow et al., 2014) (GANs)are the current state of the art machine learning data generating systems. Designed with two neural networks in the initial architecture proposal, generator and discriminator. These neural networks compete in a zero-sum game technique, to generate data having realistic properties inseparable to that of original datasets. GANs have interesting applications in various domains like Image synthesis, 3D object generation in gaming industry, fake music generation(Dong et al.), text to image synthesis and many more. Despite having a widespread application domains, GANs are popular for image data synthesis. Various architectures have been developed for image synthesis evolving from fuzzy images of digits to photorealistic images. Objectives: In this research work, we study various literature on different GAN architectures. To understand significant works done essentially to improve the GAN architectures. The primary objective of this research work is synthesis of plant images using Style GAN (Karras, Laine and Aila, 2018) variant of GAN using style transfer. The research also focuses on identifying various machine learning performance evaluation metrics that can be used to measure Style GAN model for the generated image datasets. Methods: A mixed method approach is used in this research. We review various literature work on GANs and elaborate in detail how each GAN networks are designed and how they evolved over the base architecture. We then study the style GAN (Karras, Laine and Aila, 2018a) design details. We then study related literature works on GAN model performance evaluation and measure the quality of generated image datasets. We conduct an experiment to implement the Style based GAN on leaf dataset(Kumar et al., 2012) to generate leaf images that are similar to the ground truth. We describe in detail various steps in the experiment like data collection, preprocessing, training and configuration. Also, we evaluate the performance of Style GAN training model on the leaf dataset. Results: We present the results of literature review and the conducted experiment to address the research questions. We review and elaborate various GAN architecture and their key contributions. We also review numerous qualitative and quantitative evaluation metrics to measure the performance of a GAN architecture. We then present the generated synthetic data samples from the Style based GAN learning model at various training GPU hours and the latest synthetic data sample after training for around ~8 GPU days on leafsnap dataset (Kumar et al., 2012). The results we present have a decent quality to expand the dataset for most of the tested samples. We then visualize the model performance by tensorboard graphs and an overall computational graph for the learning model. We calculate the Fréchet Inception Distance score for our leaf Style GAN and is observed to be 26.4268 (the lower the better). Conclusion: We conclude the research work with an overall review of sections in the paper. The generated fake samples are much similar to the input ground truth and appear to be convincingly realistic for a human visual judgement. However, the calculated FID score to measure the performance of the leaf StyleGAN accumulates a large value compared to that of Style GANs original celebrity HD faces image data set. We attempted to analyze the reasons for this large score.

Apprendre à éditer des images et des vidéos de visages est un domaine particulièrement actif dans la recherche académique et industrielle. Cette thèse aborde le problème de l'édition de visages dans le cas particulier des images et des vidéos à haute résolution. Dans cette thèse, nous développons des méthodes basées sur l'apprentissage profond pour effectuer l'édition d'images faciales. Plus précisément, nous explorons la tâche en utilisant les représentations latentes obtenues à partir de deux types de réseaux neuronaux profonds : les modèles basés sur l'auto-encodage et les réseaux antagonistes génératifs (GAN). Pour chaque type de méthode, nous considérons un problème spécifique d'édition d'image et proposons une solution efficace qui surpasse l'état de l'art. La thèse comprend deux parties. Dans la partie I, nous explorons les tâches d'édition d'images via l'espace latent des autoencodeurs. Nous considérons d'abord la tâche de transfert de style entre les photos, et proposons un algorithme efficace qui est construit sur une paire de réseaux basés sur des autoencodeurs. Ensuite, nous étudions la tâche d'édition de l'âge du visage pour les images à haute résolution, en utilisant une architecture d'encodeur-décodeur. Le réseau proposé encode une image de visage en représentations de caractéristiques invariantes selon l'âge, et apprend un vecteur de modulation correspondant à un âge cible. Notre approche permet une édition fine de l'âge sur des images à haute résolution dans un seul modèle unifié.Dans la deuxième partie, nous explorons la tâche d'édition via l'espace latent des modèles antagonistes génératifs (GAN). Tout d'abord, nous considérons le problème de l'édition "démêlée" (disentangled) des attributs faciaux sur des images synthétiques et réelles, en proposant un réseau de transformation latent qui agit dans l'espace latent d'un modèle GAN pré-entraîné. Nous avons également proposé un pipeline de manipulation vidéo, afin de généraliser le résultat de l'édition aux vidéos. Deuxièmement, nous étudions le problème de l'inversion du GAN - la projection d'une image réelle dans l'espace latent d'un GAN pré-entraîné. En particulier, nous proposons un encodeur feed-forward, qui encode une image donnée en un code caractéristique et un code latent en une seule passe. L'encodeur proposé s'avère plus précis et plus stable pour l'inversion d'images et de vidéos, tout en conservant de bonnes capacités d'édition
Learning to edit facial images and videos is one of the most popular tasks in both academia and industrial research. This thesis addresses the problem of face editing for the special case of high-resolution images and videos.In this thesis, we develop deep learning-based methods to perform facial image editing. Specifically, we explore the task using the latent representations obtained from two types of deep neural networks: autoencoder-based models and generative adversarial networks. For each type of method, we consider a specific image editing problem and propose an effective solution that outperforms the state-of-the-art.The thesis contains two parts. In part I, we explore image editing tasks via the latent space of autoencoders. We first consider the style transfer task between photos and propose an effective algorithm that is built on a pair of autoencoder-based networks. Second, we study the face age editing task for high-resolution images, using an encoder-decoder architecture. The proposed network encodes a face image to age-invariant feature representations and learns a modulation vector corresponding to a target age. Our approach allows for fine-grained age editing on high-resolution images in a single unified model.In part II, we explore the editing task via the latent space of generative adversarial models (GANs). First, we consider the problem of facial attribute disentangled editing on synthetic and real images, by proposing a latent transformation network that acts in the latent space of a pre-trained GAN model. We also proposed a video manipulation pipeline, to generalize the editing result to videos. Second, we investigate the problem of GAN inversion -- the projection of a real image to the latent space of a pretrained GAN. In particular, we propose a feed-forward encoder, which encodes a given image to a feature code and a latent code in one pass. The proposed encoder is shown to be more accurate and stable for image and video inversion, meanwhile, maintaining good editing capacities

Le style cognitif est un domaine bien étudié et approfondi dans les pays occidentaux, surtout aux États-Unis, mais il reste peu développé par les chercheurs chinois. Cette notion est à propos de la personnalité, de la façon de traiter des informations reçus chez les différents individus, et elle peut aussi porter une influence sur l’éducation, qui contient l’éducation de mathématiques. En Chine, les mathématiques sont une matière principale et importante dans le système éducatif, et sous un contexte particulier de la société chinoise, un système éducatif chinois s’est formé. La phase secondaire est la phase la plus représentative dans le système éducatif chinois, et sous le contexte particulier, il faut considérer les styles cognitifs les plus fréquents chez les élèves et aussi le style cognitif de chaque élève pour trouver des méthodes d’améliorer l’éducation mathématique aux écoles secondaires. Pour ce but, il faut bien prendre en considération le côté d’apprentissage et le côté de didactique. Ainsi, notre travail va commencer par une étude des recherches et des théories matures de style cognitif qui serviront de base théorique, et ensuite avec la présentation du contexte de l’éducation mathématique en Chine nous allons pouvoir mieux le lier avec le style cognitif. Avec les résultats de notre recherche nous analysons et proposons des manières possibles au niveau de politique, de didactique et d’apprentissage pour améliorer l’éducation mathématique aux écoles secondaires en profitant des théories de style cognitif et des résultats de notre recherche
The cognitive style is a well-developed and studied domain in western countries, especially in USA, but it is still a new ground for Chinese researchers. This conception is about the personality, the way to treat the received information for individuals, and it can also influence the education, which includes of course the mathematic education. In China, the mathematic is a principle subject at school in the Chinese educational system. With a special background in the society of China, a special educational system has been formed. The secondary education is the most typical in this educational system, and we need to think about the most frequent cognitive styles among Chinese students and also the cognitive style of each student to find the best methods to make the mathematic education better and more appropriate. For this destination, we need to consider the side of learning and the side of teaching. Therefore, our work will start by a presentation of the researches of cognitive styles and those mature theories, which can be our theoretical base, and then we will connect the education of mathematic with cognitive styles by a presentation of the situation of mathematics education in China. With the results of our research, we will analysis and make some advises about the politics of education, of the teaching and learning methods to try to ameliorate the mathematic education in China with help of cognitive styles and this research

Thesis advisor: Richard McGowan
Gambling, both in the casino-style and lottery forms, has risen to become a major component of the entertainment industry in the United States. State governments are the gatekeepers of this growing industry, holding the power to legalize and regulate all aspects of gambling. This thesis explores the rationale state governments have for legalizing gambling as well as the impact gambling tax revenues have for state budgets. The main focus is casino-style gambling, as casino-style gambling in particular is being pursued for expansion by numerous states in a variety of forms. As various forms of gambling are legalized throughout the country, a state's gambling interests begin to face competition from both neighboring states and other forms of gambling within the state. Econometric models attempted to predict the tax revenues a state can obtain from legalized gambling based on such competition and a states own demographics. The results support a first-mover advantage for states expanding casino-style gambling and finds that new forms of gambling significantly erode established gambling industries
Thesis (BS) — Boston College, 2006
Submitted to: Boston College. College of Arts and Sciences
Discipline: Economics Honors Program

This thesis examines the evolution of feminist praxis within the riot grrrl movement, focusing on two specific riot grrrl demographics - founding riot grrrls in the early 1990s and currently active riot grrrls in southern California. This thesis argues that riot grrrl activism is still thriving but in diverse, strategically modified ways. Using José Muñoz’s concept of “disidentification,” it analyzes how contemporary riot grrrls have appropriated and adapted the original movement’s tenets to allow for greater accessibility and diversity.

Thesis (Ph. D.)--Georgia State University, 2006.
Title from title screen. Catherine Y. Chang, committee chair; Kenneth B. Matheny, Roy M. Kern, Catherine J. Brack, Brain J. Dew, committee members. Electronic text (74 p.) : digital, PDF file. Description based on contents viewed May 29, 2007. Includes bibliographical references (p. 66-74).

Using external information to gain insights is becoming increasingly important in today’s knowledge based economies. However, from both research and practice it is somewhat unclear how organizations manage these activities. Therefore the purpose of this paper is to explore how four different organizations manage their insight creation process from external information. To do this, a literature review about insight creation is done to create an analytical framework. Based on this framework, the insight creation process is analyzed within the organizations to find characteristics of each process. Next, a cross case analysis is conducted that reveal the main differences between the organizations. The results indicate two main styles of managing the insight creation process that can be differentiated with regards to formality. As such, this paper complements previous research within the area of competitive intelligence and knowledge management in the sense that it increases the understanding of the insight creation process.

The research activity was designed to explore the idea of a predominance of one of two teaching styles, defined as teacher-centered or learner-centered, among adult basic education and adult secondary education teachers in Missouri. The Principles of Adult Learning Scale (Conti, 1985), consisting of 44 questions, was employed to identify the teaching style of respondents. The scores on the survey were compared to the educational outcome measure provided by the Department of Education and State of Missouri with the intention that conclusions as to the efficacy of one of the two styles would be revealed.

The survey was distributed to the 36 adult education programs in Missouri Three total attempts to obtain responses were made during the survey period. Of the 756 full- and part-time teachers in the Missouri system, 89 surveys were returned but due to improperly identified or unidentified numbers, only 34 of the survey responses were deemed usable.

Requiring a minimum sample of approximately 250 upon which to draw inferential conclusions, no generalizations could be drawn about the larger population of Missouri adult basic education and adult secondary education teachers. Descriptive statistics relative to the 34 participants revealed that most of the teachers were female and the highest education level was the doctorate, but most teachers held masters. The sample group average age was above 40 with 42 % older than 60. As to tenure in adult education 62 % of responding teachers had taught adult education for more than five years and 42% greater than 10 years. The 34 teachers favored teacher-centered instructional methods versus student-centered instructional methods.

Made available in DSpace on 2014-07-29T15:29:20Z (GMT). No. of bitstreams: 1 Dissertacao de mestrado barbara miranda ferreira.pdf: 429667 bytes, checksum: d95226fbacb3328ed858bfd9c2fdc8cd (MD5) Previous issue date: 2010-02-12
The excessive gestational weight gain can cause harmful consequences for both the mother and for the fetus. Nutritional, lifestyle, economic and social factors can affect the weight gain above recommendations. Objectives: To evaluate the nutritional (caloric intake, macro and micronutrients), lifestyle (smoking, physical activity, reproductive behavior) and economic (income) factors associated with the excessive weight gain in eutrophic pregnant women. Methods: This case-control study included 200 low-risk pregnant women in 2 public obstetric hospitals in Goiás, Brazil. The women that agreed to this study were divided in 2 groups: cases with excessive weight gain (> 16Kg) and controls within the recommended weight gain for eutrophic pregnant (11.50 < a < 16Kg), matched on age. The women were inquired about their food intake during the gestational period, by means of a quantitative and qualitative food-frequency questionnaire validated and adapted to the low income regional people. The women were inquired about their economic and lifestyle conditions, too. The software for statistical analyze were the SPSS version 17.0. Were used the folowing tests: Kolmogorov-Smirnov, Fisher, Student´s t, chi-square (x²), odds ratio (CI 95%), Mann-Whitney, Spearman´s correlation coefficient and multivariate linear regression. Results: The median age was 25.00 (22.00-29.00) years-old for cases and 24.00 (22.00-28.00) years-old for controls (p=0.52). The median per capita income was R$ 303.75 (225.60-500.00) for cases and R$ 333.33 (175.00-480.00) for controls (p=0.41). The median education level was 10.50 (9.00-12.00) years for cases and 11.00 (9.00-12.00) years for controls (p=0.77). There was 13.00 (13%) smokers among cases and 8.00 (8%) among controls (p=0.25). The frequency of regular physical activity was 28.00 (28%) pregnant women among cases and 31.00 (31%) among controls (p=0.65). About the reproductive behavior, there were no differences between groups for the parity (p=0.45) and abortion (p=0.90). The birth interval was 27.00 (0.00-60.75) months in cases and 22.50 (0.00-47.75) in controls (p=0.34). The following variables showed in the univariate analysis a significant correlation coefficient (ρ) to the weight gain: calories = 0.15 (p=0.04), protein = 0.16 (p=0.03), total cholesterol = 0.21 (p<0.01), retinol = 0.21 (p<0.01), iron = 0.18 (p=0.01), zinc = 0.15 (p=0.04), selenium = 0.19 (p<0.01) and phosphorus = 0.16 (p=0.03). The median of the satured fat intake was 36.57 grams (23.37 51.06) in the case group and 28.21 grams (19.92 41.90) in the control group (p=0.04). In the multivariate analysis the protein intake was the only independente variable associated with excessive pregnant weight gain. Conclusion: The factors associated with the excessive weight gain in the group of cases were total caloric intake, proteins, satured fat, cholesterol, retinol, iron, zinc, selenium and phosphorus, and in special, the protein intake. Lifestyle and economic factors didn t affect weight gain.
O ganho ponderal excessivo na gestação pode trazer consequências maléficas tanto para a mãe quanto para o feto. Fatores sociais, econômicos, de estilo de vida e nutricionais podem ser responsáveis por este ganho de peso acima das recomendações. Objetivos: Avaliar os fatores econômicos (renda), de estilo de vida (tabagismo, prática de atividade física, comportamento reprodutivo) e nutricionais (consumo calórico, macro e micronutrientes) que podem estar relacionados ao ganho de peso excessivo em gestantes eutróficas. Metodologia: Este estudo de caso-controle incluiu 200 gestantes de baixo risco obstétrico de 2 maternidades públicas do Estado de Goiás. Todas as gestantes que concordaram em participar do estudo foram divididas em 2 grupos: casos com ganho de peso excessivo (> 16Kg) e controles com ganho de peso dentro do recomendado (11,50 < a < 16Kg), pareados por idade. As mulheres foram entrevistadas com relação ao seu consumo alimentar durante o período gestacional através de um questionário quantitativo e qualitativo de frequência de consumo alimentar validado e adaptado à população regional de baixa renda. Elas também foram questionadas sobre sua condição econômica e de estilo de vida. A análise estatística foi realizada através do programa SPSS, versão 17.0. Foram aplicados os testes de Kolmogorov-Smirnov, Fisher, t de Student, qui-quadrado (x²), odds ratio (IC 95%), Mann-Whitney, coeficiente de correlação de Spearman e regressão linear multivariada. Resultados: A mediana da idade das pacientes foi 25,00 (22,00-29,00) anos para os casos e 24,00 (22,00-28,00) anos para os controles (p=0,52). A renda per capita foi de R$ 303,75 (225,60-500,00) para casos e R$ 333,33 (175,00-480,00) para controles (p=0,41). A mediana de escolaridade foi de 10,50 (9,00-12,00) anos para casos e 11,00 (9,00-12,00) anos para controles (p=0,77). Houve 13,00 (13%) fumantes no grupo de casos e 8,00 (8%) no grupo de controles (p=0,25). A frequência da prática de atividade física regular foi de 28,00 (28%) das gestantes nos casos e 31,00 (31%) nos controles (p=0,65). Em relação ao comportamento reprodutivo, não houve diferenças entre a frequência de paridade (p=0,45) e abortos (p=0,90) entre os grupos. O intervalo intergestacional foi de 27,00 (0,00-60,75) meses entre os casos e de 22,50 (0,00-47,75) meses entre os controles (p=0,34). As seguintes variáveis apresentaram na análise univariada um coeficiente de correlação (ρ) com o ganho de peso significante: calorias = 0,15 (p=0,04), proteínas = 0,16 (p=0,03), colesterol total = 0,21 (p<0,01), retinol = 0,21 (p<0,01), ferro = 0,18 (p=0,01), zinco = 0,15 (p=0,04), selênio = 0,19 (p<0,01) e fósforo = 0,16 (p=0,03). A mediana do consumo de gordura saturada foi de 36,57 gramas (23,37 - 51,06) no grupo de casos e 28,21 gramas (19,92 - 41,90) no grupo controle (p=0,04). Na análise multivariada o consumo protéico foi a única variável independente relacionada ao ganho de peso excessivo na gestação. Conclusão: O consumo calórico, as proteínas, a gordura saturada, o colesterol, o retinol, ferro, zinco, selênio e fósforo estiveram relacionados ao maior ganho ponderal observado no grupo de casos, em especial o consumo protéico. Os fatores econômicos e de estilo de vida não influenciaram o ganho de peso.

The achievement gap stands as one of the top priorities framing educational policy through the past half-century. The middle school level amplifies this gap especially in urban areas. The role of principal leadership in closing the achievement gap is key. The purpose of this study was to examine the relationships between the leadership styles of principals assigned to Title I middle schools, staff longevity, school climate, and overall school achievement. The researcher applied a non-experimental, ex-post facto research design to investigate the research hypotheses. Utilizing the Google Survey Platform, 290 staff members across 30 middle schools within a large urban school district in southeast Florida, completed a survey which included questions related to longevity, and the Multifactor Leadership Questionnaire (MLQ Form 5x). Results of the school district’s School Climate Survey, Staff Form, were employed to gauge school climate. Hierarchical linear regression analyses revealed several significant positive associations between transformational leadership and numerous of the academic areas explored (e.g., reading learning gains of the lowest quartile, math proficiency, and social studies proficiency). In contrast to what was predicted, transactional leadership, also positively predicted some of the academic achievement factors evaluated (e.g., math proficiency, and math learning gains). Staff longevity negatively predicted school grade. Staff climate positively predicted math proficiency, and math learning gains. On the other hand, staff climate negatively predicted school grade, the three factors related to reading achievement (e.g., proficiency, learning gains, and learning gains of the lowest quartile), and social studies proficiency. New research questions arose as a result of the investigation. Further research is recommended that examines the leadership variables explored within a larger sample, and in other geographical areas with similar demographics. As well, additional research is suggested involving staff longevity and school climate alongside a measure of collective instructional efficacy where urban schools are concerned.

It was not that long ago, just in the first half on the 1990s, when mobile phones were first introduced, being big and expensive. All you could do with them was to make phone calls. Since then mobile devices have experienced a great technological advance: we carry smartphones in our pockets that provide Internet access, having accelerometers that can measure acceleration, a gyroscope that can provide orientation information, different wireless interfaces such as Bluetooth connections, and above all, great computing power. On the other hand, the automobile industry has evolved significantly during the last 10 years. One of the most exciting advances in vehicle development is vehicle-to-vehicle V2V communication, which allows cars to communicate with each other over a dedicated Wi-Fi band, and share information about vehicle speed, route direction, traffic flow, and road and weather conditions. An example of such a system is GM's (General Motors) OnStar, introduced in 1996, and that provides automatic response in case of an accident, stolen-vehicle recovery, remote door unlock, and vehicle diagnostics. Also, the standard On Board Diagnosis (OBD-II), available for several years, allows us to connect to the Electronic Control Unit (ECU) via a Bluetooth OBD-II connector. This connection interface allows connectivity between the smartphone and the vehicle, and can be purchased for just over 15 euros. The spectrum of possibilities that arise when combining the car and the smartphone is unlimited, such as performing the diagnosis of the car by assuming the tasks performed by the car's On Board Unit (OBU), or sending the collected data to a platform where the diagnosis or maintenance of the system can be realized in order to detect possible faults, help you to save gas and reduce environment pollution, and notify you of your car's problems, among other features. The general objective pursued with this doctoral thesis is to help drivers to correct bad habits in their driving. To achieve this we promote the combination between smartphones and vehicular networks to design and develop a platform able to offer useful tips to achieve safer driving and greater fuel economy. It is well-known that intelligent driving can lead to lower fuel consumption, with the consequent positive impact on the environment. The proposal that has been carried out in this doctoral thesis begins with the data capture from the vehicles' OBD-II port and data analysis through the use of graphs, maps, and statistics, both, on the server itself and in the smartphone's application developed. We applied data mining techniques and neural networks to analyze, study and generate a classiffication on driving styles based on the analysis of the characteristics of each specific route used for testing. In a second phase, we demostrate the relationship between fuel consumption and driving style. To achieve that goal, the first thing that we had to realize was how to apply different algorithms for the instantaneous consumption calculation (this parameter cannot be obtained directly from the vehicle ECU). Later, we studied and analyzed all data that was collected from the drivers who shared their monitored data with the server. Although drivers do not recognize themselves as being in a state of anxiety while driving, they are more stressed than in any other daily activity, for example, when trying to stay in the right lane, keeping the car at a certain speed, and starting and stopping the vehicle. In general, drivers are more concentrated than they think, which causes an increase in the heart rate. Many factors influence heart rate while at rest, e.g. stress, medications, medical conditions, even genes play a role. In our study we also investigate how stress and the driving behavior influence the heart rate. So, in the last phase, we demostrate the correlation between heart rate and driving style, showing how the driving style can make the heart rate vary by 3 %.
No hace mucho tiempo, tan sólo en la primera mitad en la década de los 90, cuando los teléfonos móviles aparecieron, eran grandes y caros, todo lo que se podía hacer con ellos era realizar llamadas telefónicas. Desde entonces los dispositivos móviles han experimentado un gran avance tecnológico, llevamos teléfonos inteligentes en el bolsillo con acceso a Internet, acelerómetros que calculan la aceleración instantánea, giroscopios que proporcionan información de orientación, diferentes conexiones inalámbricas como Bluetooth, y sobre todo, gran capacidad de computación. Por otro lado, la industria del automóvil ha evolucionado mucho durante los últimos 10 años. Uno de los avances más interesantes en el desarrollo de vehículos ha sido la conectividad, V2V, o comunicación vehículo a vehículo, permite a los automóviles comunicarse mediante Wi-Fi y compartir información sobre la velocidad del vehículo, la dirección de la ruta actual, el tráfico, así como las condiciones de la carretera y las condiciones ambientales. Por otra parte, el estándar On Board Diagnosis (OBD-II), disponible desde hace varios años, permite conectarnos de forma sencilla a la ECU (Electronic Control Unit) mediante un conector Bluetooth OBD-II. Este interfaz de conexión permite la conectividad entre el dispositivo móvil y el vehículo, se puede adquirir por poco más de 15 euros. El espectro de posibilidades que surgen al combinar el automóvil y el Smartphone es amplísimo, como por ejemplo realizar el diagnóstico del coche a través del móvil asumiendo las tareas que hace la unidad On Board Unit (OBU) del coche, o bien enviar los datos recogidos a una plataforma donde se pueda realizar el diagnóstico o mantenimiento del sistema, detectando posibles fallos puede ayudar a ahorrar en el consumo de combustible, notificar los problemas del coche en tiempo real, entre otras características. El objetivo general que se persigue con esta tesis doctoral es ayudar al conductor a corregir malos hábitos en su forma de conducción. Conseguimos esto mediante la combinación entre smartphones y las redes vehiculares, diseñamos y desarrollamos una plataforma capaz de ofrecer consejos útiles para conseguir una conducción más segura y un mayor ahorro de combustible. Es conocido que una conducción inteligente puede llevarnos a un menor consumo de combustible, con el consiguiente impacto positivo que ello conlleva sobre el medio ambiente. La propuesta que se ha llevado a cabo en esta tesis doctoral comienza con la obtención de los datos desde el OBD-II del coche y su presentación y análisis mediante el uso de gráficas, mapas, estadísticas, tanto en el propio servidor como en la aplicación móvil desarrollada para la obtención de datos recibidos desde la ECU. Se aplicaron técnicas de minería de datos y redes neuronales para analizar, estudiar y generar una clasificación sobre los estilos de conducción en base al análisis de las características de la vía sobre la que ha realizado la ruta. En una segunda fase se demostró la relación entre el consumo de combustible con el estilo de conducción, para ello lo primero que tuvimos que realizar fue aplicar diversos algoritmos para el cálculo del consumo instantáneo, este parámetro no es posible obtenerlo directamente de la ECU del vehículo. Posteriormente se realizó el estudio y el análisis de todos los datos que se recogieron de los conductores que se prestaron a la realización del estudio enviando los datos al servidor. Muchos factores influyen en la frecuencia cardíaca en reposo, por ejemplo, el estrés, los medicamentos, las condiciones médicas, incluso los genes tienen su influencia, el envejecimiento tiende a acelerarlo, y el ejercicio regular tiende a ralentizarlo. En nuestro estudio también investigamos cómo el estrés y el comportamiento en la conducción influyen en la frecuencia cardíaca. En la última fase vemos la correlación existente entre el ri
No fa molt de temps, tan sols en la primera mitat en la dècada dels 90, quan els telèfons mòbils van aparéixer, eren grans i cars, tot el que es podia fer amb ells era realitzar telefonades. Des de llavors els dispositius mòbils han experimentat un gran avanç tecnològic, portem telèfons intel_ligents en la butxaca amb accés a Internet, acceleròmetres que calculen l'acceleració instantània, giroscopis que proporcionen informació d'orientació, diferents connexions sense _ls com Bluetooth, i sobretot gran capacitat de computació. D'altra banda, la indústria de l'automòbil ha evolucionat molt durant els últims 10 anys. Un dels avanços més interessants en el desenrotllament de vehicles ha sigut la connectivitat, V2V, o comunicació vehicle a vehicle, permet als automòbils comunicar-se per mitjà de la banda de Wi-Fi i compartir información sobre la velocitat del vehicle, la direcció de la ruta actual, les condicions del trà_c, així com l'estat de la carretera i les condicions ambientals. D'altra banda l'estàndard On Board Diagnosi (OBD-II), disponible des de fa diversos anys, permet connectar-nos de forma senzilla a l'ECU (Electronic Control Unit) per mitjà d'un connector Bluetooth OBD-II. Esta interfície de connexió permet la connectivitat entre el dispositiu mòbil i el vehicle, es pot adquirir per poc més de 15 euros. L'espectre de possibilitats que sorgixen al combinar l'automòbil i el Smartphone és il_limitat, com per exemple realitzar el diagnòstic del cotxe a través del móvil assumint les tasques que fa la unitat On Board Unit (OBU) del cotxe, o bé enviar les dades arreplegades a una plataforma on es puga realitzar el diagnòstic o manteniment del sistema, detectant possibles fallades, ajuda a estalviar en el consum de combustible, noti_car els problemes del cotxe en temps real, entre altres característiques. L'objectiu general que es perseguix amb esta tesi doctoral és ajudar al conductor a corregir mals hàbits en la seua forma de conducció. Aconseguim açò mitjançant de la combinació entre smartphones i les xarxes vehiculares, dissenyem i desenrotllem una plataforma capaç d'oferir consells útils per a aconseguir una conducció més segura i un major estalvi de combustible. És conegut que una conducció intel_ligent pot emportar-nos a un menor consum de combustible, amb el consegüent impacte positiu que això comporta sobre el medi ambient. La proposta que s'ha dut a terme en esta tesi doctoral comença amb l'obtenció de les dades des de l'OBD-II del cotxe i la seua presentació i anàlisi per mitjà de l'ús de grà_ques, mapes, estadístiques, tant en el propi servidor, com en l'aplicació mòbil desenrotllada per a l'obtenció de dades rebudes des de l'ECU. S'apliquen tècniques de mineria de dades i xarxes neuronals per a analitzar, estudiar i generar una classi_cació sobre els estils de conducció basant-se en l'anàlisi de les característiques de la via sobre la qual ha realitzat la ruta. En una segona fase es va a demostrar la relació entre el consum de combustible amb l'estil de conducció, per a això la primera cosa que vam haver de realizar va ser aplicar diversos algorismes per al càlcul del consum instantani, este paràmetre no és possible obtindre-ho directament de l'ECU del vehicle. Posteriorment es va realitzar l'estudi i l'anàlisi de totes les dades que es van arreplegar dels conductors que es van prestar a la realització de l'estudi enviant les dades al servidor. Molts factors in_ueixen en la freqüència cardíaca en repòs, per exemple, l'estrès, els medicaments, les condicions mèdiques, _ns i tot els gens tenen la seua in_uència, l'envelliment tendeix a accelerar-ho, i l'exercici regular tendeix a ralentir-ho. En el nostre estudi només estem interessats en com l'estrès i el comportament en la conducció in_ueixen en la freqüència cardíaca. En l'última fase vam veure la correlació existent entre el ritme cardíac i l'estil de conducci
Meseguer Anastasio, JE. (2017). A novel framework to promote eco-driving through smartphone-vehicle integration [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/84287
TESIS

碩士
國立中興大學
台灣文學與跨國文化研究所
103
This dissertation is trying to study the locality, history and style in Gan Yao-Ming’s Ghost Slayer. Gan’s literary outputs remove traditional constraints, and are totally different from the traditional local literatures which emphasize grief and sorrow, as well as abandon all heavy missions, to create the brand all new writing field of the local novel lightly. The point of this dissertation is discussing how Gan uses the magic of language to describe the locality and how he reappears the heavy history in his pen, furthermore, how he creates his own particular essay style in novel to express esthetics. Ghost Slayer follows the line of grand history, depicts in various and intersected ways to exhibit overlapped phenomenon. Firstly, we’ll discuss local imaginations in this novel. Native place is named GuanNiuWo, just the hometown of Gan, a Hakka village. Since this, we can tell that the locality of the novel is also Hakka locality. This novel apply the different aspects of Native Hakka language to invest the Hakka manner, method and emotion. The author blended local and Hakka language together and express the finest of Taiwaneses. Secondly, we analyze the field of history in Ghost Slayer to emphasize the concept of localization through Ghost King and local legend, train and intermixing identification in colonial history to discover the ethnic code in Hakka angle. And further, we’ll probe into the esthetic genre from Gan’s complicated language character and the writing technique as well as temple. Finally, we’ll explore story telling skill and style. Gan makes readers feel postponed atmosphere, as his poetic words premeat unique presentation of realiy. Like a magician, he has created a fantastic epic of history with profound concern and deeply care about the local.

This dissertation tackles the problem of entanglement in Generative Adversarial Networks (GANs). The key insight is that disentanglement in GANs can be improved by di erentiating between the content, and the operations performed on that content. For example, the identity of a generated face can be thought of as the content, while the lighting conditions can be thought of as the operations. We examine disentan- glement in several kinds of deep networks. We examine image-to-image translation GANs, unconditional GANs, and sketch extraction networks. The task in image-to-image translation GANs is to translate images from one domain to another. It is immediately clear that disentanglement is necessary in this case. The network must maintain the core contents of the image while changing the stylistic appearance to match the target domain. We propose latent  lter scaling to achieve multimodality and disentanglement. Previous methods require complicated network architectures to enforce that disentanglement. Our approach, on the other hand, maintains the traditional GAN loss with a minor change in architecture. Unlike image-to-image GANs, unconditional GANs are generally entangled. Un- conditional GANs o er one method of changing the generated output which is chang- ing the input noise code. Therefore, it is very di cult to resample only some parts of the generated images. We propose structured noise injection to achieve disentan- glement in unconditional GANs. We propose using two input codes: one to specify spatially-variable details, and one to specify spatially-invariable details. In addition to the ability to change content and style independently, it also allows users to change the content only at certain locations. Combining our previous  ndings, we improve the performance of sketch-to-image translation networks. A crucial problem is how to correct input sketches before feeding them to the generator. By extracting sketches in an unsupervised way only from the spatially-variable branch of the image, we are able to produce sketches that show the content in many di erent styles. Those sketches can serve as a dataset to train a sketch-to-image translation GAN.

Esta tese é um estudo de caso sobre a problemática das culturas juvenis em contexto urbano e a produção das suas sociabilidades e estilos de vida. O seu trabalho de campo circunscreveu-se à Freguesia da Arrentela (Seixal), pano de fundo da acção colectiva do grupo de jovens Red Eyes Gang. Maioritariamente descendentes de imigrantes africanos, os integrantes desta crew tentam construir identidades positivas (ao recusar os discursos que os representam como “feios, porcos e maus”), num contexto marcado pela pobreza e estigmatização. Este grupo informal apropria-se do estilo rap com uma postura que visa pôr em causa as noções dominantes sobre o seu lugar social, contribuindo na construção de novos significados sobre a sua identidade enquanto jovens, pobres e negros. Ao mesmo tempo, o estilo rap serve de resistência e demarcação aos valores oficiais da sociedade e desenvolve alternativas de inserção social num contexto marcado pela fragilidade das instituições do Estado. Embora os jovens Red Eyes Gang partilhem identificações e estilos de vida, estes adquirem contornos múltiplos. Se há questões importante que os unificam – caso contrário não faria sentido a existência deste colectivo –, há outras que os diferenciam e que estão em disputa no seu interior. Por isso, há diferentes formas de ser jovem e de se apropriar do estilo rap no interior do grupo, o que contraria os discursos que apresentam os jovens das classes desfavorecidas sob uma óptica homogénea e pouco dinâmica.
The present thesis is a case study of youth cultures in urban settings and of their socialization strategies and life styles. The field work was carried out in the borough of Arrentela (Seixal), which provides the background to the collective action of the Red Eyes Gang. The members of this crew, which are mostly descendants of African immigrants, try to create positive identities (by refusing the discourses that represent them as “ugly, dirty and mean”) in a setting characterized by poverty and stigmatization. This informal group uses rap in a way as to challenge the dominant notions about their social roles, contributing to the production of new meanings to their identity as youngsters, poor and black. Rap serves as a means of resistance to and demarcation from the official social standards and at the same time develops social insertion alternatives in a setting characterized by the weakness of State institutions. Even though the Red Eyes Gang members share identities and life styles, these acquire different shapes. While there are important issues that unite them – otherwise the existence of the group would make no sense – others differentiate them and are in dispute. Therefore, there are different ways of being young and of using rap within the group, which is in contradiction to the discourses that portray lower class youngsters from a homogeneous and very little dynamic perspective.

This dissertation is a study of the comic intermezzo’s literary origins and musical practice in the years before Pergolesi’s La serva padrona (1733). It begins with a chronological examination of Italian comic plays and operas written between 1660 and 1723. During these years comic playwrights adopted a style of writing speech from the improvised theatre which makes use of what Richard Andrews (1993) refers to as “elastic gags.” This style of comedy flourished under Medici patronage in Florence in the last decades of the seventeenth century and then spread to Venice, Rome and Naples during the first years of the intermezzo’s development. It is a style of comedy shared with the plays of Molière, and other contemporaneous French authors. This dissertation examines several scenes based on French works which have previously not been identified as having earlier sources. The decision to adapt these earlier sources for the intermezzo did not occur in a vacuum. The practice of comedy in the intermezzo was conditioned by the artistic, social and political climate of Italy. This study investigates the relationship between intermezzos and the milieus which produced them. The success of some intermezzos, like Il marito giocatore (1719), resulted from a combination of their artistic merit and their broad social appeal, while others, like Albino e Plautilla (1723), were musically adept but remained obscure because their humour was specific to the world they satirized. Both intermezzos are indebted to earlier French sources. Many others which are metatheatrical in nature draw on contemporary debates about opera. A final section examines selected arias from the intermezzo repertory using incongruity theory. Comic theory makes clear that the intermezzo’s musical language was not a new development. Just as librettists drew on earlier written traditions to form the literary text of the intermezzo, composers drew on existing musical practices to create humour. The intermezzo was therefore not naively comic—a portrait of the genre which is all too common—but rather a repertory which was thoroughly enmeshed within contemporary artistic practice and a wider social and cultural world.

碩士
大葉大學
管理學院碩士在職專班
102
This study aimed to investigate the current status of students’ learning attitude and learning satisfaction with the implementation of new aerobics, and probe into how different research variables affect students’ learning attitude and learning satisfaction with aerobics. It also intended to understand the correlation between students’ learning attitude and learning satisfaction. This study enrolled the students in Shih Gang Junior High School in Taichung City in academic year 2012 as the research subjects. A total of 356 questionnaires were distributed, and 356 questionnaires were returned. After eliminating the 40 invalid samples, there were 316 valid samples. The questionnaire content included students’ background information, learning attitude scale, and learning satisfaction scale. The data were analyzed with SPSS 12.0 for descriptive statistics, t test, One-way ANOVA, Scheffé post hoc comparison, and Pearson product-moment correlation. The results of the statistical analyses on the data were as follows: 1.In general, the current status of students’ learning attitude and learning satisfaction with new aerobics was neutral to agree (satisfied). In the dimension of learning attitude, they were most satisfied with “skill,” and the score of “affection” was the lowest. In the dimension of learning satisfaction, they were most satisfied with “teachers’ teaching,” and the score of “content of teaching material” was the lowest. 2.In terms of the effect of different background variables on the learning attitude and learning satisfaction with new aerobics, “gender” and “grade” both had a significant effect on them. However, “parent education” did not have a significant effect on the learning attitude towards and learning satisfaction with new aerobics. 3.There was a positive correlation between learning attitude and learning satisfaction with new aerobics. Finally, based on the research results, this study proposed suggestions as the reference for competent education authority, junior high schools, teachers and future studies.

Musical timbre transfer is the task of re-rendering the musical content of a given source using the rendering style of a target sound. The source keeps its musical content, e.g., pitch, microtiming, orchestration, and syncopation. I specifically focus on the task of transferring the style of percussive patterns extracted from polyphonic audio using a MelGAN-VC model [57] by training acoustic properties for each genre. Evaluating audio style transfer is challenging and typically requires user studies. An analytical methodology based on supervised and unsupervised learning including visualization for evaluating musical timbre transfer is proposed. The proposed methodology is used to evaluate the MelGAN-VC model for musical timbre transfer of drum tracks. The method uses audio features to analyze results of the timbre transfer based on classification probability from Random Forest classifier. And K-means algorithm can classify unlabeled instances using audio features and style-transformed results are visualized by t-SNE dimensionality reduction technique, which is helpful for interpreting relations between musical genres and comparing results from the Random Forest classifier.
Graduate

Running title: Personality and cyberbullying victimisation
Includes bibliographical references (leaves 104-120)
The aim of this study was to gain insight into the relationship between personality and cyberbullying victimisation using a sample of 107 adult participants. The participants were found using the social media site Facebook in particular and assessed by means of the Big Five Inventory and a cyberbullying measure developed by the researcher. The study also explored the relationship between personality and coping behaviours using the COPE Inventory. The study findings indicate that neuroticism and cyberbullying victimisation are correlated, albeit weakly, with the largest correlation being between neuroticism and online impersonation and harassment. In addition, the study found a weak correlation between openness to experience and online impersonation. Coping and personality were moderately correlated; with neurotic individuals and victims of cyberbullying often resorting to maladaptive coping strategies. These results may suggest that there are other, more significant risk factors in the cyberbullying field that merit further exploration, while the relationships highlighted in the study also require further investigation.
Psychology
M.A. (Psychology)

Purpose: The strategy-to-performance gap is a contemporary problem that causes organisations to perform less than optimal or fail. Leadership, strategy and performance are complex areas of research on the topic with lack of conclusive results, and solutions to the issue remain elusive. This study offered a clearer understanding of the problem. Design: This qualitative study explored leadership style in the context of strategy execution with the objective of offering a better understanding of performance and the strategy-to-performance gap. The study is a case study of selected organisations in the wine industry in South Africa. Data were collected via interviews on leadership figures and employees. Data were analysed qualitatively. Findings: The study confirmed the existence and issues surrounding the strategy-to-performance gap. Some findings concurred with previous studies on the gap and confirmed that it is intertwined with leadership and strategy execution issues. New themes were identified that contributed to the field of study and could prove to be valuable to narrow the strategy-to-performance gap in future. Value: Leadership is linked with performance in literature but there was limited research found on leadership style specifically in the context of the strategy-to-performance gap. This study confirmed issues surrounding the gap and it was found that the issue is a contemporary problem affecting organisational performance in various ways. This contributed to a better understanding of the problem. New themes that emerged could prove valuable to further research. Leadership and specifically leadership styles (as per Lewin) were found to influence and be interwoven with the strategy-to-performance gap. This study is original as it is the first study that explored leadership styles specifically in the context of the other key themes: strategy and the strategy-to-performance gap. The new themes that emerged are leadership style requirements from an employee satisfaction perspective as well as specific skills suggested to improve performance, all highlighted in the findings. Recommendations: Further research could explore leadership styles in more detail with regard to the strategy-to-performance gap. New themes in the context of strategy execution and leadership style could also be valuable. Future studies could also potentially evaluate the key themes with other measurement instruments such as different leadership styles or different performance criteria
Business Management
M. Com. (Business Management)