No description available.

Submitted by PPG Medicina e Ci?ncias da Sa?de (medicina-pg@pucrs.br) on 2018-07-13T12:02:18Z No. of bitstreams: 1 DAVID_PONCIANO_DE_SENA.pdf: 7136289 bytes, checksum: 63e5ded0ad929d80daf354b846f5dd28 (MD5) / Approved for entry into archive by Sheila Dias (sheila.dias@pucrs.br) on 2018-07-18T12:59:17Z (GMT) No. of bitstreams: 1 DAVID_PONCIANO_DE_SENA.pdf: 7136289 bytes, checksum: 63e5ded0ad929d80daf354b846f5dd28 (MD5) / Made available in DSpace on 2018-07-18T13:04:40Z (GMT). No. of bitstreams: 1 DAVID_PONCIANO_DE_SENA.pdf: 7136289 bytes, checksum: 63e5ded0ad929d80daf354b846f5dd28 (MD5) Previous issue date: 2018-04-12 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior - CAPES / Aim of the Study: The purpose of this study was to develop and validate a serious game for mobile platforms comparing with a video-assisted self-learning method to assist in the teaching and learning process of cardiopulmonary resuscitation (CPR) maneuvers. In a traditional training, CPR is based on learning by doing using a simulated model with the assistance of an instructor. However, this model presents an overall higher cost and less accessibility, provoking the need for cheaper and more accessible alternative methods. Methods: Forty-five volunteer first-year medical students completed a written multiple choice and practical pretest about CPR maneuvers skills and were randomly allocated into two groups. During a period of 20 minutes, the video self-learning group with 22 students was exposed to a video-based training about CPR, while the video game group with 23 students used a serious game simulating a cardiac arrest scenario where the student should perform virtual CPR to keep playing the game. Each group then performed, a written multiple choice score and practical post-test on a CPR training model while being evaluated by three blinded emergency doctors based on 2015 AHA-BLS (American Heart Association - Basic Life Support) protocol. Both groups were also evaluated about how long they kept interested on each self-learning system. Results: The video group had superior performance as confirmed by a written multiple choice score 7.56+-0.21 against 6.51+-0,21 for a video game (p=0. 001) and practical scores 9.67+-0.21 against 8.40+-0,21 for a video game (p<0. 001). The video game group stayed longer using the method as confirmed by 18.57+- 0,66 minutes for video game group and 7.41+-0,43 for the video group (p<0. 001), demonstrating greater interest in the video game method. Conclusions: The group that used a video game as a self-training method in a short period of exposure had a lower performance than the video group in both the theoretical and practical tests regarding cardiorespiratory resuscitation. However, there was a clear preference for students to use games rather than videos as a form of self-training. / Objetivo do estudo: O objetivo deste estudo foi desenvolver e validar um jogo s?rio para plataformas m?veis em compara??o com um m?todo de autoaprendizagem assistido por v?deo para auxiliar no processo de ensino e aprendizagem das manobras de ressuscita??o cardiopulmonar (RCP). Em um treinamento tradicional, a RCP ? baseada no aprendizado pela pr?tica, utilizando um modelo simulado com a ajuda de um instrutor. No entanto, este modelo apresenta um custo global mais elevado e de menor acessibilidade, despertando a necessidade de m?todos alternativos mais baratos e pratic?veis. M?todos: Quarenta e cinco volunt?rios, estudantes de medicina do primeiro ano, completaram um pr?-teste de m?ltipla escolha escrito e um pr?-teste pr?tico sobre RCP e foram alocados aleatoriamente em dois grupos. Durante um per?odo de 20 minutos, o grupo de autoaprendizagem v?deo, composto por 22 alunos, foi exposto a um v?deo de treinamento sobre a RCP, enquanto o grupo videogame, composto por 23 estudantes, utilizou um jogo s?rio, simulando um cen?rio de parada card?aca, onde o aluno deveria executar uma RCP virtual para continuar jogando. Ao t?rmino do treinamento, cada grupo foi submetido a um p?s-teste escrito de m?ltipla escolha e um p?s-teste pr?tico em um modelo de treinamento de RCP, sendo avaliados de forma cegada, por tr?s m?dicos com experi?ncia em atendimento de emerg?ncia, com base no protocolo 2015 AHA-BLS (American Heart Association - Basic Life Support). Ambos os grupos tamb?m foram avaliados quanto a quantidade de tempo que eles permaneciam interessados em cada sistema de autoaprendizagem. Resultados: O grupo v?deo apresentou desempenho superior confirmado por uma maior pontua??o no teste escrito de m?ltipla escolha, 7.56 + -0.21 contra 6.51 + -0,21 para videogame (p = 0. 001) e pontua??o no teste de avalia??o pr?tica 9,67 + -0,21 contra 8,40 + -0,21 para videogame (p <0. 001). O grupo videogame permaneceu mais tempo utilizando o m?todo, 18,57 + - 0,66 minutos para o grupo videogame e 7,41 + -0,43 para o grupo v?deo (p <0. 001), demonstrando maior interesse no m?todo do videogame. Conclus?es: O grupo que usou o jogo s?rio (grupo videogame) como um m?todo de autotreinamento em um curto per?odo de exposi??o teve um desempenho pior do que o grupo v?deo nos testes te?ricos e pr?ticos em rela??o ? ressuscita??o cardiopulmonar. No entanto, houve uma clara prefer?ncia por parte dos alunos em utilizar jogos em rela??o aos v?deos como forma de autotreinamento.

Cyberbullying has become a significant societal concern as people increasingly use computing technology to mediate all elements of their life. Cyberbullying can cause serious psychological and emotional problems for people who are impacted. Hence, developing automated ways for detecting cyberbullying is critical. While recent efforts to identify cyberbullying have specified advanced text processing approaches, the area of visual data processing has received significantly less attention.  This thesis presents multimodal approach on cyberbullying detection in social media by using a combination of visual and textual features. The work presents the prototype of cyberbullying detection system alongside its design and implementation details.  The foundings of this work prove that visual features can be used for the improvement of the textual approach. The results show that visual features is a best performing approach according to Naive Bayes and Random Forest algorithms and combination of textual and visual features reached the highest performance according to Support Vector Classification model. Furthermore, the result reveal that feature combination leading to the highest performance of Support Vector Classification is a combination of text description, text sentiment and image tags. The strongest feature performance with Random Forest and Naive Bayes model has been demonstrated by image tags. / Cybermobbning har blivit ett stort samhällsproblem i takt med att människor i allt högre grad använder datorteknik i alla delar av sitt liv. Cybermobbning kan orsaka allvarliga psykologiska och känslomässiga problem för de personer som drabbas. Därför är det viktigt att utveckla automatiserade metoder för att upptäcka cybermobbning. Medan de senaste insatserna för att identifiera cybermobbning har specificerat avancerade metoder för textbearbetning, har visuell databehandling fått betydligt mindre uppmärksamhet.  I den här avhandlingen presenteras en multimodal metod för att upptäcka cybermobbning i sociala medier med hjälp av en kombination av visuella och textuella element. I arbetet presenteras prototypen av ett system för upptäckt av cybermobbning tillsammans med detaljer om dess utformning och genomförande.  Resultaten av detta arbete visar att visuella funktioner kan användas för att förbättra den textuella modellen. Resultaten visar att visuella funktioner är det bästa tillvägagångssättet enligt Naive Bayes - och Random Forest­algoritmerna och att kombinationen av text- och visuella funktioner uppnådde den högsta prestandan enligt Support Vector Classification - modellen. Dessutom visar resultatet att kombinationen av funktioner som leder till den högsta prestandan för Support Vector Classification ar en kombination av textbeskrivning, textkänsla och bildtaggar. Den starkaste prestandan med Random Forest - och Naive Bayes -modellen däremot, har visats med bildtaggar.

No description available.

<p dir="ltr">This thesis presents probably approximately correct (PAC) bounds on estimates of the transition kernels of Controlled Markov chains (CMC’s). CMC’s are a natural choice for modelling various industrial and medical processes, and are also relevant to reinforcement learning (RL). Learning the transition dynamics of CMC’s in a sample efficient manner is an important question that is open. This thesis aims to close this gap in knowledge in literature.</p><p dir="ltr">In Chapter 2, we lay the groundwork for later chapters by introducing the relevant concepts and defining the requisite terms. The two subsequent chapters focus on non-parametric estimation. </p><p dir="ltr">In Chapter 3, we restrict ourselves to a finitely supported CMC with d states and k controls and produce a general theory for minimax sample complexity of estimating the transition matrices.</p><p dir="ltr">In Chapter 4 we demonstrate the applicability of this theory by using it to recover the sample complexities of various controlled Markov chains, as well as RL problems.</p><p dir="ltr">In Chapter 5 we move to a continuous state and action spaces with compact supports. We produce a robust, non-parametric test to find the best histogram based estimator of the transition density; effectively reducing the problem into one of model selection based on empricial processes.</p><p dir="ltr">Finally, in Chapter 6 we move to a parametric and Bayesian regime, and restrict ourselves to Markov chains. Under this setting we provide a PAC-Bayes bound for estimating model parameters under tempered posteriors.</p>

This study aimed to assess the effectiveness of an instructor-led and a self-managed computer training course. A sample of forty nine (n=49) previously disadvantaged South African adult learners was used. Half the sample was randomly assigned to the instructor-led course, while the other half was assigned to the self-managed course. Data dealing with the course content and design, subject demographics, previous computer experience, preferred learning style, and learning potential was collected prior to each course. After the course, the ability to create key outcomes using the learned software was assessed, as well as subject perceptions of the course and various support and performance system factors. The results showed that there was no significant difference between the performance on the competence assessment of the students from the two groups. The data did, however, indicate a stronger preference for the self-managed approach. A number of limitations to the study were also noted. / Industrial and Organisational Psychology / M.A. (Industrial Psychology)

This study aimed to assess the effectiveness of an instructor-led and a self-managed computer training course. A sample of forty nine (n=49) previously disadvantaged South African adult learners was used. Half the sample was randomly assigned to the instructor-led course, while the other half was assigned to the self-managed course. Data dealing with the course content and design, subject demographics, previous computer experience, preferred learning style, and learning potential was collected prior to each course. After the course, the ability to create key outcomes using the learned software was assessed, as well as subject perceptions of the course and various support and performance system factors. The results showed that there was no significant difference between the performance on the competence assessment of the students from the two groups. The data did, however, indicate a stronger preference for the self-managed approach. A number of limitations to the study were also noted. / Industrial and Organisational Psychology / M.A. (Industrial Psychology)