This thesis deals by natural disasters and solutions through commercial insurance. The introduction explains the concept of natural disaster and subsequently there are closer specified individual natural hazards. Then, the work focuses on the development of catastrophic events in the world in recent years, the consequences of their formation, their effects and the impact on these events on the world and Czech insurance. Due to the high frequency of floods in the Czech Republic in recent years, the work also addresses the issue of floods, their insurability and changes on the Czech insurance market reacting to the increased occurrence of floods. The thesis focuses on commercial property insurance products of individuals given to individual areas of different risk in terms of flooding. The analysis shows that insurers evaluate the individual risk zones in different ways and the most risk area of insurance excluded. At the end, there are evaluate the scope of insurance risk transfer to other entities, its importance and possibilities of alternative risk transfer.

abstract: In the last decade deep learning based models have revolutionized machine learning and computer vision applications. However, these models are data-hungry and training them is a time-consuming process. In addition, when deep neural networks are updated to augment their prediction space with new data, they run into the problem of catastrophic forgetting, where the model forgets previously learned knowledge as it overfits to the newly available data. Incremental learning algorithms enable deep neural networks to prevent catastrophic forgetting by retaining knowledge of previously observed data while also learning from newly available data. This thesis presents three models for incremental learning; (i) Design of an algorithm for generative incremental learning using a pre-trained deep neural network classifier; (ii) Development of a hashing based clustering algorithm for efficient incremental learning; (iii) Design of a student-teacher coupled neural network to distill knowledge for incremental learning. The proposed algorithms were evaluated using popular vision datasets for classification tasks. The thesis concludes with a discussion about the feasibility of using these techniques to transfer information between networks and also for incremental learning applications. / Dissertation/Thesis / Masters Thesis Computer Science 2020

archives@tulane.edu / 1 / Kimberly Michelle Koporc

No description available.

Generalized collaborative drones are a technology that has many potential benefits. General purpose drones that can handle exploration, navigation, manipulation, and more without having to be reprogrammed would be an immense breakthrough for usability and adoption of the technology. The ability to develop these multi-task, multi-agent drone systems is limited by the lack of available training environments, as well as deficiencies of multi-task learning due to a phenomenon known as catastrophic forgetting. In this thesis, we present a set of simulation environments for exploring the abilities of multi-task drone systems and provide a platform for testing agents in incremental single-agent and multi-agent learning scenarios. The multi-task platform is an extension of an existing drone simulation environment written in Python using the PyBullet Physics Simulation Engine, with these environments incorporated. Using this platform, we present an analysis of Incremental Learning and detail the beneficial impacts of using the technique for multi-task learning, with respect to multi-task learning speed and catastrophic forgetting. Finally, we introduce a novel algorithm, Incremental Learning with Second-Order Approximation Regularization (IL-SOAR), to mitigate some of the effects of catastrophic forgetting in multi-task learning. We show the impact of this method and contrast the performance relative to a multi-agent multi-task approach using a centralized policy sharing algorithm. / Master of Science / Machine Learning techniques allow drones to be trained to achieve tasks which are otherwise time-consuming or difficult. The goal of this thesis is to facilitate the work of creating these complex drone machine learning systems by exploring Reinforcement Learning (RL), a field of machine learning which involves learning the correct actions to take through experience. Currently, RL methods are effective in the design of drones which are able to solve one particular task. The next step in this technology is to develop RL systems which are able to handle generalization and perform well across multiple tasks. In this thesis, simulation environments for drones to learn complex tasks are created, and algorithms which are able to train drones in multiple hard tasks are developed and tested. We explore the benefits of using a specific multi-task training technique known as Incremental Learning. Additionally, we consider one of the prohibitive factors of multi-task machine learning-based solutions, the degradation problem of agent performance on previously learned tasks, known as catastrophic forgetting. We create an algorithm that aims to prevent the impact of forgetting when training drones sequentially on new tasks. We contrast this approach with a multi-agent solution, where multiple drones learn simultaneously across the tasks.

Language modeling, especially through the use of transformer-based large language models (LLMs), has drastically changed how we view and use artificial intelligence (AI) and machine learning (ML) in our daily lives. Although LLMs have showcased remarkable linguistic proficiency in their abilities to write, summarize, and phrase, these model have yet to achieve the same remarkability in their ability to quantitatively reason. This deficiency is specially apparent in smaller models (less than 1 Billion parameters) than can run natively on-device. Between the complementary capabilities of qualitative and quantitative reasoning, this thesis focuses on the latter, where the goal is to devise mechanisms to instill quantitative reasoning capabilities into these models. However, instilling this notion is not as straight forward as traditional end-to-end learning. The learning of quantitative notions include the ability of the model to discern between regular linguistic tokens and magnitude/scale-oriented non-linguistic tokens. The learning of these notions, specially after pre-training, comes at a cost for these models: catastrophic forgetting. Thus, learning needs to be followed with retention - making sure these models do not forget what they have learned. Thus, we first motivate the need for numeracy-enhanced models via their potential applications in field of data-to-text generation (D2T), showcasing how these models behave as quantitative reasoners as-is. Then, we devise both token-level training interventions and information-theoretic training interventions to numerically enhance these models, with the latter specifically focused on combating catastrophic forgetting. Our information-theoretic interventions not only lead to numerically-enhanced models but lend us critical insights into the learning behavior of these models, especially when it comes to adapting these models to the target task distribution from their pretraining distribution. Finally, we extrapolate these insights to devise more effective strategies transfer learning and unlearning for language modeling. / Doctor of Philosophy / Language modeling, especially through the use of transformer-based large language models (LLMs), has drastically changed how we view and use artificial intelligence (AI) and machine learning (ML) in our daily lives. Although LLMs have showcased remarkable linguistic proficiency in their abilities to write, summarize, and phrase, these model have yet to achieve the same remarkability in their ability to quantitatively reason. This deficiency is specially apparent in smaller models than can run natively on-device. This thesis focuses on instilling within these models the ability to perform quantitative reasoning - the ability to differentiate between words and numbers and understand the notions of magnitude tied with said numbers, while retaining their linguistic skills. The learned insights from our experiments are further used to devise models that better adapt to target tasks.

Background: Non-communicable diseases (NCDs) are the leading causes of morbidity and mortality among adults in Vietnam. Little is known about the magnitude of socioeconomic inequalities in NCDs and other NCD-related factors in urban areas, in particular among the poor living in slum areas. Understanding these disparities are essential in contributing to the knowledge, needed to reduce inequalities and close the related health gaps burdening the disadvantaged populations in urban areas.  Objective: To examine the burden and health system responsiveness to NCDs in Hanoi, Vietnam and investigate the role of socioeconomic inequalities in their prevalence, subsequent healthcare utilization and related impoverishment due to health expenditures.  Methods: A cross-sectional study was conducted among 3,736 individuals aged 15 years and over who lived in 1211 randomly selected households in 2013 in urban Hanoi, Vietnam. The study collected information on household’s characteristics, household expenditures, and household member information. A qualitative approach was implemented to explore the responsiveness of commune health stations to the increasing burden of NCDs in urban Hanoi. In-depth interview approach was conducted among health staff involved in NCD tasks at four commune health stations in urban Hanoi. Furthermore, NCD managers at relevance district, provincial and national levels were interviewed.  Results: The prevalence of self-reported NCDs was significantly higher among individuals in non-slum areas (11.6%) than those in slum areas (7.9%). However, the prevalence of self-reported NCDs concentrated among the poor in both slum and non-slum areas. In slum areas, the poor needed more health care services, but the rich consumed more health care services. Among households with at least one household member reporting diagnosis of NCDs, the proportion of household facing catastrophic health expenditure and impoverishment were the greater in slum areas than in non-slum areas. Poor households in slum areas were more likely to face catastrophic health expenditure and impoverishment. The poor in non-slum areas were also more likely to face impoverishment if their household members experienced NCDs. Health system responses to NCDs at commune health stations in urban Hanoi were weak, characterized by the lack of health information, inadequate human resources, poor financing, inadequate quality and quantity of services, lack of essential medicines. The commune health stations were not prepared to respond to the rising prevalence of NCDs in urban Hanoi.  Conclusion: This thesis shows the existence of socioeconomic inequalities in the prevalence of self-reported NCDs in both non-slum and slum areas in urban Hanoi. NCDs associated with the inequalities in health care utilization, catastrophic health expenditure and impoverishment, particular in slum areas. Appropriate interventions should focus more on specific population groups to reduce the socioeconomic inequalities in the NCD prevalence and health care utilization related to NCDs to prevent catastrophic health expenditure and impoverishment among the households of NCD patients.  The functions of commune health stations in the urban setting should be strengthened through the development of NCDs service packages covered by the health insurance.

A quercetina é um flavonoide com alta atividade antioxidante reconhecida. Devido a este fato, a indústria alimentícia tem buscado alternativas para a aplicação deste bioativo como antioxidante em matrizes alimentícias. No entanto, incorporar a quercetina em formulações alimentícias pode ser desafiador, pois sua molécula tem relativo grau de hidrofobicidade. Os métodos de encapsulação em nanoemulsões podem tornar viável a proteção do flavonoide, bem como permitir sua dispersibilidade em meios aquosos . O objetivo do presente trabalho foi produzir nanoemulsões (fase oleosa: óleo de girassol) encapsulando quercetina pelo método do ponto de inversão da emulsão (EIP, emulsion inversion point), determinando parâmetros de operação que viabilizem um futuro escalonamento do processo, e a realização de um estudo de caso sobre a incorporação das nanodispersões em patê de frango. Os parâmetros de produção que foram avaliados foram : tipo e concentração de tensoativo (razão SOR, razão tensoativo:óleo), concentração de óleo, concentração de cosolvente e velocidade de agitação. Foram produzidas nanoemulsões utilizando dois tensoativos diferentes, Tween 80 e Brij 30. As porcentagens de quercetina encapsulada foram 0,15 e 0,30% (m/m). Dentre estas, a concentração de 0,30% foi a mais viável para a aplicação no patê de frango, por apresentar melhor distribuição de tamanho de gotícula (diâmetro hidrodinâmico), não apresentar variação significativa de oxidação lipídica e maior conservação da quercetina encapsulada por um período de 90 dias. Os valores de tensão interfacial obtidos mostram que a quercetina provoca abaixamento da tensão interfacial do sistema, o que pode indicar que a quercetina forma um complexo com os tensoativos utilizados, fato que influencia decisivamente na sua localização nas nanogotas. Das formulações testadas para patê de frango contendo quercetina, apenas a formulação usando quercetina livre não apresentou boa avaliação nos itens cor e sabor. Tal resultado justifica a necessidade da incorporação da quercetina em nanoemulsões para não causar alterações sensoriais no produto. / Quercetin is a flavonoid with recognized high antioxidant activity Due to this fact, the food industry has been trying to use it as an alternative as a preservative in food matrices. However, incorporating quercetin into food formulations may be challenging due to its relative degree of hydrophobicity. Encapsulation methods in nanoemulsions may render the protection of the flavonoid viable, as well as enable its dispersibility in aqueous media. The objective of the present study was to produce nanoemulsions (oil phase: sunflower oil) encapsulating quercetin by the emulsion inversion point method (EIP), determining operating parameters that enable a future process scale-up. A case study on the incorporation of nanodispersions into chicken patê was also carried out. The production parameters evaluated were: type and concentration of surfactant (SOR ratio, surfactant:oil ratio), soybean oil concentration, cosolvent concentration and stirring speed. Nanoemulsions were produced using two different surfactants, Tween 80 and Brij 30. The percentages of encapsulated quercetin were 0.15 and 0.30% (m / m). Among these, the concentration of 0.30% was the most feasible for application in chicken pâté, since it presented better droplet size distribution (hydrodynamic diameter), did not present significant variation of lipid oxidation and higher capacity of preserving encapsulated quercetina for a period of 90 days. The interfacial tension values obtained show that quercetin lowered the interfacial tension of the system, which may indicate that quercetin forms a complex with the surfactants used, a fact that decisively influences its location in the nanodroplets. Among the formulations tested in chicken pâté containing quercetin, only the formulation using free quercetin did not show good acceptance evaluation on the color and taste itens. Such an evaluation was important as it justifies the need for the incorporation of quercetin into nanoemulsions in order to not cause significant sensorial changes in the product.

Submitted by JOSIANE SANTOS DE OLIVEIRA (josianeso) on 2017-05-23T12:22:56Z No. of bitstreams: 1 Thiago Feltes Marques_.pdf: 1461587 bytes, checksum: dc52bcfe3a1e36c05d530f5b912ce8e8 (MD5) / Made available in DSpace on 2017-05-23T12:22:56Z (GMT). No. of bitstreams: 1 Thiago Feltes Marques_.pdf: 1461587 bytes, checksum: dc52bcfe3a1e36c05d530f5b912ce8e8 (MD5) Previous issue date: 2016-12-21 / Nenhuma / O presente trabalho estuda a aplicação do Princípio da Precaução diante dos riscos catastróficos. Os Desastres ambientais possuem potencial magnitude e complexidade, razão pela qual vem sendo desenvolvido no Brasil um ramo do Direito voltado aos riscos e danos catastróficos. Ocorre que, por se tratar de um ramo diferenciado do Direito Ambiental, seus componentes também possuem novas feições. Um deles é o Princípio da Precaução, conceituado no Direito Ambiental como um poderoso mecanismo para lidar com a incerteza científica diante da ameaça de danos sérios ou irreversíveis e ausência de absoluta certeza científica, devendo ser irradiado para concretizar o processo de tomada de decisões no seio da administração do risco, condicionado à observância dos limites estabelecidos pelas normas da proporcionalidade, da não-discriminação (igualdade), da coerência (razoabilidade), conforme os parâmetros da evolução científica de cada local e época. Diante de riscos catastróficos, contudo, o Princípio da Precaução adquire uma forma refinada, pois tendo em vista suas graduações e intensidades diversas onde, quanto maior a expectativa de gravidade (verificado pela Avaliação de Impacto Ambiental), este deve ser mais restritivo. A precaução para catástrofes deve ser mais sensível às informações científicas, pois a magnitude do desastre é muito intensa. Por essa razão, a presente dissertação preocupou-se em observar quais as formas de gestão dos riscos catastróficos, tendo como base as dimensões da incerteza, assim como pretendeu-se aclarar a difícil e complexa questão com os aportes do referencial teórico especializado em Desastres e Princípio da Precaução, permitindo consequentemente enxergar a forma de aplicação da baliza com a lente da antecipação aos riscos catastróficos e seus danos. / The present paper studies the application of the Precautionary Principle in the face of the catastrophic risks. Environmental disasters have potential impact and complexity, which is why Brazil has developed a branch of law focused on risks and catastrophic damages. Whathever, because it’s a branch differentiated from Environmental Law, his components also have new features. One of them is the Precautionary Principle, which is conceptualized in Environmental Law as a powerful mechanism to deal with scientific uncertainty in the face of the threat of serious or irreversible damage and lack of absolute scientific certainty, and must be irradiated to concretize the decision-maker process within the risk management, subject to compliance with the limits established by the rules of proportionality, no-discrimination (equality), consistency (reasonableness), according to the parameters of the scientific evolution of each place and time. In the face of catastrophic risks, however, the Precautionary Principle acquires a refined form, since in view of graduations and differents intensities, the greater the expect of gravity (verified by the Environmental Impact Assessment), the latter must be more restrictive. Caution for catastrophes should be more sensitive to scientific information because the impact of the disaster is very intense. For this reason, the present work was concerned with the management of catastrophic risks, based on the dimensions of uncertainty, on clarifying the difficult and complex issue with the contributions of the theoric framework specialized in Disasters and Precautionary Principle, allowing to see the way of applying the principle with the lens of antecipation to the catastrophic risks and their damages.

Ce travail porte sur l’inversion de phase catastrophique induite par l’agitation continue d’un système anormal eau-huile-surfactif, sans addition de phase, afin de réaliser des émulsions eau-dans-huile concentrées et finement dispersées. Les suivis rhéologiques et conductimétriques des systèmes sous agitation indiquent que l’inversion passe, en général, par la formation d’une émulsion multiple e/H/E dans laquelle une partie de la phase continue est inclue comme gouttelettes dans les gouttes de phase dispersée. Ainsi, le volume apparent de phase dispersée augmente jusqu’à atteindre une valeur critique à laquelle l’inversion se produit. Afin de maîtriser ce processus, l’influence de variables de formulation, composition et agitation sur le mécanisme d’inversion et sur ses paramètres caractéristiques a été étudié. Les données recueillies permettent de choisir les conditions de formulation et de procédé pour préparer, en un temps minimal, des émulsions E/H avec des propriétés requises. / This study deals with the catastrophic phase inversion produce by continuous stirring of an abnormal water-oil-surfactant system, without internal phase addition, in order to prepare concentrated and fine water-in-oil emulsions. The simultaneous conductivity and viscosity measurements of the system under stirring show that generally, the inversion takes place through the formation of a multiple w/O/W emulsion in which a portion of the external phase is continuously included as droplets in the dispersed phase drops. Consequently, the dispersed phase apparent volume increases until a critical value is reached and the inversion is triggered. In order to control this process, the influence of formulation, composition and stirring variables on the inversion mechanism and on its characteristic parameters is studied. The data collected can be used to prepare, in a minimal time, emulsions with required properties, by controlling the formulation and process conditions.