Afin de répondre aux besoins de plus en plus importants en puissance de calcul de la part des applicationsnumériques, les supercalculateurs ont dû évoluer et sont ainsi de plus en plus compliqués àprogrammer. Ainsi, en plus de l’apparition des systèmes à mémoire partagée, des architectures ditesNUMA (Non Uniform Memory Access) sont présentes au sein de ces machines, fournissant plusieursniveaux de parallélisme. Une autre contrainte, la diminution de la mémoire disponible par coeur decalcul, doit être soulignée. C’est ainsi que des modèles parallèles tels que MPI (Message Passing Interface)ne permettent plus aux codes scientifiques haute performance de passer à l’echelle et d’exploiterefficacement les machines de calcul, et doivent donc être combinés avec d’autres modèles plus adaptésaux architectures à mémoire partagée. OpenMP, en tant que modèle standardisé, est un choix privilégiépour être combiné avec MPI. Mais mélanger deux modèles avec des paradigmes différents est unet âche compliquée et peut engendrer des goulets d’étranglement qui doivent être identifiés. Cette thèsea pour objectif d’aborder ces limitations et met en avant plusieurs contributions couvrant divers aspects.Notre première contribution permet de r éduire le surcoût des supports exécutifs OpenMP en optimisantle travail d’activation et de synchronisation des threads OpenMP pour les codes MPI+OpenMP. Dansun second temps, nous nous focalisons sur les opérations collectives MPI. Notre contribution a pourbut d’optimiser l’opération MPI Allreduce en réutilisant des unités de calcul inoccupées, et faisant intervenirdes threads OpenMP. Nous introduisons également le concept de collectives unifiées, impliquantdes tâches MPI et des threads OpenMP dans une même opération. Enfin, nous nous intéressons àl’analyse de performance et plus précisément l’instrumentation des applications MPI+OpenMP, et notredernière contribution consiste en l’implémentation et l’ évaluation de l’outil OpenMP Tools API (OMPT)dans le support exécutif OpenMP du framework MPC. Cet outil nous permet d’instrumenter des constructionsOpenMP et de conduire une analyse axée aussi bien du côté des applications que dessupports d’exécution / To provide increasing computational power for numerical simulations, supercomputers evolved and arenow more and more complex to program. Indeed, after the appearance of shared memory systemsemerged architectures such as NUMA (Non Uniform Memory Access) systems, providing several levelsof parallelism. Another constraint, the decreasing amount of memory per compute core, has to bementioned. Therefore, parallel models such as Message Passing Interface (MPI) are no more sufficientto enable scalability of High Performance applications, and have to be coupled with another modeladapted to shared memory architectures. OpenMP, as a de facto standard, is a good candidate to bemixed with MPI. The principle is to use this model to augment legacy codes already parallelized withMPI. But hybridizing scientific codes is a complex task, bottlenecks exist and need to be identified. Thisthesis tackles these limitations and proposes different contributions following various aspects. Our firstcontribution reduces the overhead of the OpenMP layer by optimizing the creation and synchronizationof threads for MPI+OpenMP codes. On a second time, we target MPI collective operations. Our contributionconsists in proposing a technique to exploit idle cores in order to help the operation, with theexample of MPI Allreduce collective. We also introduce unified Collectives involving both MPI tasks andOpenMP threads. Finally, we focus on performance analysis of hybrid MPI+OpenMP codes, and ourlast contribution consists in the implementation of OpenMP Tools API (OMPT), an instrumentation tool,inside the OpenMP runtime of MPC framework. This tool allows us to instrument and profile OpenMPconstructs and allows the analysis of both runtime and application sides

<p>U okviru predmetne disertacije izvršena je detaljna analiza razvijenih<br />modela za procenu i upravljanje troškovima životnog ciklusa proizvoda<br />i istaknut značaj upravljanja ovim troškovima u fazi razvoja proizvoda.<br />Životni ciklus proizvoda definisan je fazama razvoja, proizvodnje,<br />upotrebe i reciklaže, sa odgovarajućim aktivnostima, kao uzročnicima<br />troškova. Za ovako definisani životni ciklus razvijen je hibridni<br />model upravljanja troškovima životnog ciklusa u fazi razvoja<br />proizvoda. Procena troškova životnog ciklusa novog proizvoda u ovom<br />modelu vrši se primenom fazi neuronskih mreža, dok se upravljanje<br />troškovima vrši razvojem konstrukcije u fazi konceptualnog i<br />preliminarnog razvoja dizajna novog proizvoda i usvajanjem procesa<br />proizvodnje koja obezbeđuje ciljne troškove i konkurentnost proizvoda na<br />tržištu u pogledu troškova upotrebe i pogodnosti za reciklažu.</p> / <p>In the context of this thesis, a detailed analysis of the developed models for<br />assessing and managing the life cycle costs of products and highlighted the<br />importance of managing these costs in the product development phase.<br />The life cycle of a product is defined stages of development, production, use<br />and recycling, with appropriate activities as cost drivers. For such a defined<br />life cycle model developed hybrid life cycle cost management in the product<br />development phase. Assessment of life cycle costs of a new product in this<br />model is made with fuzzy neural network, while managing costs is performed<br />at the stage of construction, development of conceptual and preliminary<br />design development of new products and the adoption of production<br />processes that ensures the target costs and the competitiveness of products<br />in the market in terms of cost and use facilities for recycling.</p>

The 2008/9 financial crisis intensified the search for realistic return models, that capture real market movements. The assumed underlying statistical distribution of financial returns plays a crucial role in the evaluation of risk measures, and pricing of financial instruments. In this dissertation, we discuss an empirical study on the evaluation of the traditional portfolio risk measures, and option pricing under the hybrid Brownian motion model, developed by Shaw and Schofield. Under this model, we derive probability density functions that have a fat-tailed property, such that “25-sigma” or worse events are more probable. We then estimate Value-at-Risk (VaR) and Expected Shortfall (ES) using four equity stocks listed on the Johannesburg Stock Exchange, including the FTSE/JSE Top 40 index. We apply the historical method and Variance-Covariance method (VC) in the valuation of VaR. Under the VC method, we adopt the GARCH(1,1) model to deal with the volatility clustering phenomenon. We backtest the VaR results and discuss our findings for each probability density function. Furthermore, we apply the hybrid model to price European style options. We compare the pricing performance of the hybrid model to the classical Black-Scholes model. / Dissertation (MSc)--University of Pretoria, 2017. / National Research Fund (NRF), University of Pretoria Postgraduate bursary and the General Studentship bursary / Mathematics and Applied Mathematics / MSc / Unrestricted

Hearing loss is considered a global health concern with 466 million people affected worldwide. Current hearing health care delivery models are based on several consecutive face-to-face consultations that occur in-person. Information and communications technology, and especially mobile technology, can be used to support or enhance health care delivery. This can be employed in addition, or as an alternative to, current patient treatment pathways. This project developed a hybrid hearing health care approach by combining online and face-to-face services. The services were offered using a five-step approach: (1) online hearing screening, (2) motivational engagement by voice/video calling, messaging, or emailing, (3) diagnostic hearing testing in a face-to-face appointment, (4) counseling, hearing aid trial and fitting using face-to-face and online methods, and (5) online aural rehabilitation, counseling and ongoing coaching using face-to-face and online methods. Three studies were conducted. Study I investigated the readiness, characteristics and behaviors of patients who sought hybrid hearing health care. Over three months (June–September 2017), 462 individuals completed the online hearing screening test: 59% (271/462) of those failed (age M = 60; SD = 12), indicating that further assessment and treatment might be required. These patients had been aware of their hearing loss for a period of between 5 to 16 years. A significant positive correlation was observed between age and speech reception threshold (r = 0.21; p < .001), where older participants presented with poorer scores. Five participants completed readiness measurement scales and attended a face-to-face diagnostic hearing evaluation during this time. Study II investigated patient uptake, experience and satisfaction with hybrid hearing health care using a process evaluation. The process evaluation study was conducted over a three-month period for patients who sought services from the clinic over a period of 19 months (June 2017–January 2019). A total of 46 patients seen at the clinic were invited to complete an online questionnaire regarding their experiences and satisfaction with the steps completed and services provided. Of those, 31 (67%) patients responded (age M = 66; SD = 16). Of the 61% of patients who had previously sought hearing services, 95% reported the hybrid clinic services as superior. Two main themes emerged from the patient’s comparison of their experience with the hybrid clinic versus previous experiences: clinician engagement (personal attention, patience, dedication, thorough explanations, professional behavior, exceeding expectations, friendliness and trust) and technology (latest technology, advanced equipment and hearing aid trial). Patients who completed all five steps, including acquiring hearing aids and taking part in an online aural rehabilitation program (continued with hearing health care), were significantly older and had significantly poorer speech reception thresholds compared to those who did not acquire hearing aids after the diagnostic hearing test and hearing aid trial (discontinued hearing health care). A significant positive correlation was found between age and the number of face-to-face appointments attended per patient (r = 0.37; p = .007). Study III investigated whether digital proficiency (proficiency with mobile devices and computers) was a predictor of the uptake of hybrid hearing health care. A total of 931 individuals failed the online hearing screening test and had submitted their details to the clinic for further care over a 24-month period (June 2017–June 2019). Of the 931 online test takers, 53 persons (age M = 64; SD = 15) who attended a face-to-face diagnostic hearing testing completed a mobile device and computer proficiency questionnaire. An exact regression model identified age as the factor associated with patients completing all five steps, including acquiring hearing aids and taking part in an online aural rehabilitation program (continued with hearing health care) from a hybrid model (β = .07; p = .018). Older patients were more likely to continue to seek hearing health care. Digital proficiency was not significantly associated with adults with hearing loss taking up services through a hybrid hearing health care model. The results from these three studies demonstrate that asynchronous internet-based services such as an online hearing screening test can be used to create awareness of hearing health care. It is possible to provide online support to patients during the initial stages of seeking hearing health care online prior to the first face-to-face visit. Patient uptake, satisfaction and experience of using hybrid hearing health care services are positive when compared to traditional methods of service delivery. Hearing health care models that combine face-to-face and online methods hold promise for audiologists willing to incorporate online modalities into current treatment pathways. This research project highlights the opportunity for audiologists to provide services and personalized support to patients using a combination of face-to-face and online modalities. / Thesis (PhD)--University of Pretoria, 2020. / This work was supported by the National Research Foundation (NRF) of South Africa under the grant number 107728. / Speech-Language Pathology and Audiology / PhD (Audiology) / Unrestricted

This thesis deals with comparing two approaches to modelling and predicting time series: a traditional one (the ARIMAX model) and a modern one (gradiently boosted decision trees within the framework of the XGBoost library). In the first part of the thesis we introduce the theoretical framework of supervised learning, the ARIMAX model and gradient boosting in the context of decision trees. In the second part we fit the ARIMAX and XGBoost models which both predict a specific time series, the daily volume of the S&P 500 index, which is a crucial task in many branches. After that we compare the results of the two approaches, we describe the advantages of the XGBoost model, which presumably lead to its better results in this specific simulation study and we show the importance of hyperparameter optimization. Afterwards, we compare the practicality of the methods, especially in regards to their computational demands. In the last part of the thesis, a hybrid model theory is derived and algorithms to get the optimal hybrid model are proposed. These algorithms are then used for the mentioned prediction problem. The optimal hybrid model combines ARIMAX and XGBoost models and performs better than each of the individual models on its own. 1

A multitude of data is being produced from the increase in instructional technology, e-learning resources, and online courses. This data could be used by educators to analyze and extract useful information which could be beneficial to both instructors and students. Educational Data Mining (EDM) extracts hidden information from data contained within the educational domain. In data mining, hybrid method is the combination of various machine learning techniques. Through this dissertation, the novel use of machine learning hybrid techniques was explored in EDM using three educational case studies. First, in consideration for the importance of students’ attention, on and off-task data to analyze the attention behavior of the students were collected. Two feature selection techniques, Principal Component Analysis and Linear Discriminant Analysis, were combined to improve the classification accuracies for classifying the students’ attention patterns. The relationship between attention and learning was also studied by calculating Pearson’s correlation coefficient and p-value. Our examination was then shifted towards academic performance as it is important to ensuring a quality education. Two different 2D- Convolutional Neural Network (CNN) models were concatenated and produced a single model to predict students’ academic performance in terms of pass and fail. Lastly, the importance of using machine learning in online learning to maintain academic integrity was considered. In this work, primarily a traditional machine learning algorithms were used to predict the cheaters in an online examination. 1D CNN architecture was then used to extract the features from our cheater dataset and the previously used machine learning model was applied on extracted features to detect the cheaters. Such type of hybrid model outperformed the original traditional machine learning model and CNN model when used alone in terms of classification accuracy. The three studies reflect the use of machine learning application in EDM. Classification accuracy is important in EDM because different educational decisions are made based on the results of our model. So, to increase the accuracies, a hybrid method was employed. Thus, through this dissertation it was successfully shown that hybrid models can be used in EDM to improve the classification accuracies.

Planning, scheduling and real time optimization (RTO) are currently implemented by using different types of models, which causes discrepancies between their results. This work presents a single model of a crude distillation unit (preflash, atmospheric, and vacuum towers) suitable for all of these applications, thereby eliminating discrepancies between models used in these decision processes. Hybrid model consists of volumetric and energy balances and partial least squares model for predicting product properties. Product TBP curves are predicted from feed TBP curve, operating conditions (flows, pumparound heat duties, furnace coil outlet temperatures). Simulated plant data and model testing have been based on a rigorous distillation model, with 0.5% RMSE over a wide range of conditions. Unlike previous works, we do not assume that (i) midpoint of a product TBP curve lies on the crude distillation curve, and (ii) midpoint between the back-end and front-end of the adjacent products lies on the crude distillation curves, since these assumptions do not hold in practice. Associated properties (e.g. gravity, sulfur) are computed for each product based on its distillation curve. Model structure makes it particularly amenable for development from plant data. High model accuracy and its linearity make it suitable for optimization of production plans or schedules. / Thesis / Master of Applied Science (MASc)

This research identifies shared strengths of Scrum and Kanban, such as iterative and incremental work and effective workload management, which are particularly beneficial in startup contexts characterized by high uncertainty. By systematically comparing the fundamental elements of both methodologies, the study delineates key dimensions and components that inform the development of a hybrid approach. This study aimed to explore which components of Kanban and Scrum are best suited for navigating uncertainty and ambiguity in a fast-paced environment. To achieve this objective, our first research question delves into the productive attributes of both Scrum and Kanban methods, laying the groundwork for our investigation. Additionally, we sought to identify the key factors driving adaptations in the application of these methods within software startups, as well as pinpoint the highlighted concepts and elements that can be effectively utilized. This formed the basis of our second research question, with the ultimate aim of developing a hybrid model that integrates the most beneficial aspects of both methodologies, thus addressing the challenges posed by uncertainty and rapid change in the initial phase of the study. We employed a mixed-methods approach, utilizing a literature review, survey, and case study analysis. The theoretical foundation was established through a comprehensive review of existing literature on Scrum and Kanban, supported by the construction of a comparative mind map. Survey data was collected from industry professionals to gather insights into the practical application of these methodologies. Additionally, a case study was conducted to provide a deeper understanding of the challenges and successes in implementing agile practices within a startup environment. Through our research, we identified key dimensions and components of Scrum and Kanban relevant to software startups, including values, principles/practices, team structure, events/cadences, and artifacts. By analyzing survey responses and conducting in-depth interviews, we gained valuable insights into the strengths and limitations of each methodology. Moreover, our case study provided real-world examples of the challenges faced and strategies employed in a startup setting. The findings suggest that while Scrum and Kanban offer distinct strengths, a hybrid approach integrating both methodologies can better address the dynamic needs of software startups. The proposed hybrid model emphasizes flexibility and adaptability, allowing startups to tailor practices to their specific context while leveraging the strengths of Scrum and Kanban. Overall, this research contributes to a deeper understanding of how startups can navigate agile transformation and achieve success in a competitive market.

Objetivo. Desenvolvimento de uma estrutura para aplicação de simulação numérica estocástica no estudo de espalhamento de doenças em metapopulações de maneira que esta incorpore a topologia dinâmica de contatos entre as subpopulações, verificando as peculiaridades do modelo e aplicando este modelo às redes de movimentação de animais de Pernambuco para estudar o papel das feiras de animais. Método. Foi utilizado o paradigma de modelos híbridos para tratar do espalhamento de doenças nas metapopulações que, das nossas aplicações, resultou na união de duas estratégias de modelagem: Modelos Baseados no Indivíduo e o Algorítimo de Simulação Estocástica. Aplicamos os modelos híbridos em redes de movimentação de animais reais e fictícias para destacar as diferenças dos modelos híbridos com diferentes abordagens de migração (pendular e definitiva) e comparamos estes modelos com modelos clássicos de equações diferenciais. Ainda, através do pacote hybridModels, estudamos o papel das feiras de animais em cenários de epidemia de febre aftosa na rede de movimentação de animais de Pernambuco, introduzindo a doença numa feira de animais contida numa amostra da base de Guia de Trânsito Animal e calculamos a cadeia de infecção dos estabelecimentos. Resultados. Constatamos que no estudo de epidemias com o uso de modelo híbrido, a migração pendular, na média, subestima o número de animais infectados no cenário de comercialização de animais (migração defi nitiva), além de traduzir uma dinâmica de espalhamento enganosa, ignorando cenários mais complexo oferecido pela migração definitiva. Criamos o pacote hybridModels que generaliza os modelos híbridos com migração definitiva e com ele aplicamos um modelo híbrido SIR na rede de Pernambuco e verificamos que as feiras de animais de Pernambuco são potentes disseminadores de doenças transmissíveis. Conclusão. Apesar de custo computacional maior no estudo de espalhamento de doenças, a migração definitiva é o mais adequado tipo de conexão entre as subpopulações de animais de produção. Ainda, de acordo com as nossas analises, as feiras de animais estão entre os mais importantes nós na rede de movimentação de Pernambuco e devem ter lugar de destaque nas estratégias de controle e vigilância epidemiológica / Objective. Development of framework applied to stochastic numerical simulation for the study of disease spreading in metapopulations, in a way that it incorporates the dynamic topology of contacts between subpopulations, checking the framework peculiarities and applying it to the animal movement network of Pernambuco to study the role of animal markets. Method. We used hybrid models paradigm to treat disease spread in metapopulations. From our applications it has resulted in the union of two modeling strategies: Individual-based model and the Algorithm for Stochastic Simulation. We applied hybrid models in real and fictitious networks to highlight the differences between different animal movement approaches (commuting and migration) and we compared these models with classic models of differential equations. Furthermore, through the hybridModels package, we studied the role of animal markets in epidemic scenarios of Foot and Mouth Disease (FMD) in animal movement networks of Pernambuco, introducing the disease in an animal market of a sample from the Animal Transit Record of Pernambuco&rsquo;s database and calculating the contact infection chain of premises. Results. We noted that in the study of epidemics using a hybrid model, commuting can underestimates the number of infected animals in the animal trade scenario (migration), and resulting in a misleading spreading dynamic by ignoring a more complex scenario that occurs with migration. We created the hybridModels package that generalizes the hybrid models with migration, applied a SIR hybrid model to the animal movement network of Pernambuco and verified that animal markets are important disease spreaders. Conclusion. Despite its higher computational cost in the study of epidemics in animal movement networks, migration is the most suitable type of connection between subpopulations. Furthermore, animal markets of Pernambuco are among the most important nodes for disease transmission and should be considered in strategies of surveillance and disease control

Synaptic vesicle recycling is the reuse of synaptic membrane material and proteins after vesicles have been exocytosed at the pre-synaptic terminal of a neuronal synapse. The discovery of the mechanisms by which recycling operates is a subject of active research. Within small mammalian central nervous system nerve terminals, two studied mechanisms of recovery are clathrin-mediated endocytosis and activity-dependent bulk endocytosis. Research into the comparative kinetics and mechanisms underlying these endocytosis mechanisms commonly involves time-series fluorescence microscopy of in vitro cultures. Synaptic proteins are tagged with fluorescent markers, or the synaptic vesicles are labelled with fluorescent dye. The change in fluorescence levels of individual synapses over time in response to stimuli is used to understand synaptic activity. The image analysis of these time-series images frequently requires substantial manual effort to extract the changing synaptic fluorescence intensity levels over time. This work focusses on two closely interlinked areas, the development of improved automated image analysis tools to facilitate the analysis of microscopy image data, and computational simulations to leverage the data obtained from these experiments to gain mechanistic insight into the underlying processes involved in synaptic vesicle recycling. The imaged properties of synapses within the time-series images are characterised, in terms of synapse movement during the course of an experiment. This characterisation highlights the properties which risk adding error to the extracted fluorescence intensity data, as analysis generally requires segmentation of regions of interest with fixed size and location. Where possible, protocols to optimise the manual selection of synapses in the image are suggested. The manual selection of synapses within time-series images is a common but time consuming and difficult task. It requires considerable skill on the part of the researcher to select synapses from noisy images without introducing error or bias. Automated tools for either general image segmentation or for segmentation of synapse-like puncta do exist, but have mixed results when applied to time-series experiments. This work introduces the use of knowledge of the experiment protocol into the segmentation process. The selection of synapses as they respond to known stimuli is compared against other current segmentation methods, and tools to perform this segmentation are provided. This use of synapse activity improves the quality of the segmented set of synapses over existing segmentation tools. Finally, this work builds a number of computational models, to allow published individual data points to be aggregated into a coherent view of overall synaptic vesicle recycling. The first is FM-Sim, a stochastic hybrid model of overall synapse recycling as is expected to occur during the course of an experiment. This closed system model handles the processes of exocytosis and endocytosis. It uses Bayesian inference to fit model parameters to experimental data. In particular, it uses the experimental protocol to separate the mechanisms and rates that may contribute to the observed experimental data. The second is a mathematical model of one aspect of synaptic vesicle recycling of particular interest - homoeostasis of plasma membrane integrity on the presynaptic terminal. This model provides bounds on efficiency of the studied endocytosis mechanisms at recovery of plasma membrane area during and after neuronal stimulus. Both the image analysis and the computational simulations demonstrated in this work provide useful tools and insights into current research of synaptic vesicle recycling and the role of activity-dependent bulk endocytosis. In particular, the utility of adding time-dependent experimental protocol knowledge to both the image analysis tools and the computational simulations is shown.