Spiking neural networks have, in recent years, become a popular tool for investigating the properties and computational performance of large massively connected networks of neurons. Equally as interesting is the investigation of the potential computational power of individual spiking neurons. An overview is provided of current and relevant research into the Liquid Sate Machine, biologically inspired artificial STDP learning mechanisms and the investigation of aspects of the computational power of artificial, recurrent networks of spiking neurons. First, it is shown that, using simple structures of spiking Leaky Integrate and Fire (LIF) neurons, a network n(P), can be built to perform any program P that can be performed by a general parallel programming language. Next, a form of STDP learning with normalisation is developed, referred to as STDP + N learning. The effects of applying this STDP + N learning within recurrently connected networks of neurons is then investigated. It is shown experimentally that, in very specific circumstances Anti-Hebbian and Hebbian STDP learning may be considered to be approximately equivalent processes. A metric is then developed that can be used to measure the distance between any two spike trains. The metric is then used, along with the STDP + N learning, in an experiment to examine the capacity of a single spiking neuron that receives multiple input spike trains, to simultaneously learn many temporally precise Input/Output spike train associations. The STDP +N learning is further modified for use in recurrent networks of spiking neurons, to give the STDP + NType2 learning methodology. An experiment is devised which demonstrates that the Type 2 method of applying learning to the synapses of a recurrent network — effectively a randomly shifting locality of learning — can enable the network to learn firing patterns that the typical application of learning is unable to learn. The resulting networks could, in theory, be used to create to simple structures discussed in the first chapter of original work.

Neste trabalho será investigado o comportamento de escoamentos de fluidos newtonianos e não-newtonianos em microcanais. O problema não-newtoniano, consiste em resolver as equações que regem o movimento para o caso de um escoamento de fluidos cujas propriedades reológicas possam ser estudadas pelo modelo constitutivo de Phan-Thien-Tanner, como por exemplo os materiais poliméricos. Uma das características interessantes de alguns destes materiais é que eles podem ser misturados com solventes apropriados, como uma solução eletrolítica, e o resultado é que este fluido como um todo passa a ter propriedades elétricas. Assim, além das propriedades viscoelásticas, será investigada a eletrocinética do escoamento, que é diretamente influenciado pela aplicação de um campo elétrico externo. Em particular o fenômeno de eletrosmose será estudado neste trabalho por meio de simulações numéricas em canais planos e bocais. Escoamentos através de bocais podem ser úteis na realização da mistura de fluidos. Neste trabalho, apresentamos uma aproximação numérica para a simulação dos vórtices que ocorrem logo após a passagem do fluido pela contração no bocal. Além disso, foram feitas simulações com potenciais dependentes do tempo e as variações nas propriedades do fluido viscoelástico para este tipo de escoamento. O movimento das cargas na solução é descrito pelas equações de Poisson-Nernst-Planck e para resolver numericamente este problema será aplicado o método das diferenças finitas generalizadas. O código para as simulações de escoamentos eletrosmóticas foi implementado como uma parte do sistema chamado HiG-Fow, que é capaz de realizar simulações de outros tipos de escoamentos além dos investigados nesta tese. A modularidade desse sistema possibilita ao usuário a implementação de pacotes de acordo com o tipo de escoamento a ser estudado. Os resultados obtidos indicam que a tensão viscoelástica é resistente a mudanças bruscas no escoamento, isto é, as oscilações da tensão normal dependem da frequência aplicada. Quanto maior a frequência menor é a amplitude do tensor. Esperamos que este trabalho contribua para um melhor entendimento de escoamentos dependente do tempo, possibilitando melhorias na tecnologia de dispositivos microfluidicos. / In this work the behavior of newtonian and non-Newtonian fluids in microchannels will be investigated. The non-Newtonian problem consists in solving the governing equations of the movement take into account a fluid flow whose rheological properties can be studied by the Phan- Thien-Tanner constitutive model, for example the polymeric materials. One of the interesting features of some of these materials is that they can be mixed with appropriate solvents, such as an electrolyte solution, and the resulting fluid has electrical properties. Thus, besides the viscoelastic properties, the electrokinetics of the flow will be investigated, which is directly influenced by the application of an external electric field. In particular the phenomenon of electro-osmosis will be studied in this work through numerical simulations in flat channels and nozzles. Flows through nozzles may be useful in performing the fluid mixture. In this work, we propose a numerical approximation for the simulation of the vortices that occur soon after the passage of the fluid by the contraction in the nozzle. In addition, simulations were made with time dependent potentials and the variations in viscoelastic fluid properties for this type of flow. The motion of the charges in the solution is described by the Poisson-Nernst-Planck equations and to solve numerically this problem will be applied the generalized finite difference method. The code for the simulations of electroosmotic flows was implemented as a part of the HiG-Fow system, which is able to perform simulations of other types of flows beyond those investigated in this thesis. The modularity of this system enables the user to implement packages according to the type of flow to be studied. The results indicate that the viscoelastic stress is resistant to abrupt changes on the flow, that is, the oscillations of the normal stress depend on the applied frequency. The higher the frequency the smaller the tensor amplitude. We hope that this work contributes to a better understanding of time-dependent flows, enabling improvements in the technology of microfluidic devices.

Orientador: Liciana Vaz de Arruda Silveira / Resumo: Nas pesquisas clínicas, o método usado para analisar os fatores que contribuem para a mortalidade relacionada aos mais diversos tipos de desfecho é a análise de sobrevivência. Na pesquisa clínica há mais de um resultado possível durante o acompanhamento dos dados de sobrevivência, estes são conhecidos como eventos competitivos. Na ocorrência dos eventos em que a morte é considerada um evento censurado, o modelo de risco proporcional de Cox pode ser empregado para estimar os efeitos das covariáveis sobre o risco. Entretanto, o efeito sobre o risco não pode ser diretamente afetado pela função de incidência acumulada (FIA). Assim, de maneira geral, o presente estudo teve por objetivo fazer uma abordagem dos métodos de análise de sobrevida nos aspectos paramétricos de modelos flexíveis para riscos competitivos, e aplicá-los a uma coorte do Sul da Itália, para avaliar os riscos de óbito para câncer gastrointestinal e outras causas. Trata-se de um estudo observacional retrospectivo, com um período de seguimento de dez anos, realizado em duas coortes, totalizando 2.623 indivíduos selecionados segundo a lista de eleitores dos municípios Castellana Grotte e Putgnano, na Itália. Os modelos flexíveis para riscos competitivos foram usados para avaliar as distribuições acumuladas de incidência específica por causa. Observou-se que os indivíduos com câncer gastrointestinal adeptos da dieta mediterrânea apresentam um risco constante e menor de morrer, especialmente nas etapas inicial e fina... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: In clinical research, the method used to analyze the factors that contribute to mortality related to the most diverse types of outcome is the survival analysis. In clinical research there is more than one possible outcome during monitoring of survival data, these are known as competitive events. In the occurrence of events in which death is considered a censored event, the Cox proportional hazard model can be used to estimate the effects of covariates on risk. However, the effect on risk can not be directly affected by the cumulative incidence function (FIA). Thus, in general, the present study aimed to analyze the methods of survival analysis in the parametric aspects of flexible models for competitive risks, and to apply them to a cohort in the South of Italy to evaluate the risk of death for gastrointestinal cancer and other causes. This is a retrospective observational study with a follow-up period of ten years, carried out in two cohorts, totaling 2,623 individuals selected according to the list of voters of the municipalities Castellana Grotte and Putgnano, Italy. Flexible models for competitive risks were used to evaluate cumulative distributions xiii of specific incidence by cause. It was observed that individuals with gastrointestinal cancer adherents of the Mediterranean diet present a constant and lower risk of dying, especially in the initial and final stages of the follow-up, whereas this effect is less pronounced in individuals who did not adhere to the aforemen... (Complete abstract click electronic access below) / Doutor

Endogenous viral elements, or viral genomic fossils, have proven extremely valuable in the study of the macroevolution of viruses, providing important, and otherwise unobtainable, insights into the ancient origin of viruses, and how their ancestors might have co-evolved with their hosts in the distant past. This type of investigation falls within the realm of paleovirology—the study of ancient viruses. Investigations of extant viruses and paleovirological analyses, however, often give conflicting results, especially those concerning viral evolutionary rates and timescales. Reconciling these two types of analyses is a necessary step towards a better understanding of the overall long-term evolutionary dynamics of viruses. The main study system of this thesis is foamy viruses (FVs). FVs are characterised by their stable co-speciation history with their hosts, allowing their evolutionary dynamics to be modelled and investigated over various timescales. This unique evolutionary feature makes FVs one of the best subjects for connecting recent and ancient viral evolution. The work here reports the discovery of several endogenous mammalian FVs, and examines how mammalian FVs co-evolve with their hosts. Analyses reveal a co-diversifying history of the two that could be dated back to the basal radiation of eutherians more than 100 million years ago. However, a small number of ancient FV cross-species transmissions could still be found, mostly involving New World monkey FVs. Based on this extended FV-mammal co-speciation pattern, this thesis investigates the long-term evolutionary rate dynamics of FVs, and shows that the rate estimates of FV evolution appear to decrease continuously as the rate measurement timescale increases, following a power-law decay function. The work presented here also shows that this so-called 'time-dependent rate phenomenon' is in fact a pervasive evolutionary feature of all viruses, and surprisingly, the rate estimates of evolution of all viruses seem to decay at the same speed, decreasing by approximately half for every 3-fold increase in the measurement timescale. Based on this power-law rate-decay pattern, we could infer evolutionary timescales of modern-day lentiviruses that are consistent with paleovirological analyses for the first time. Finally, this thesis reports the discovery of basal FV-like endogenous retroviruses (FLERVs) in amphibian and fish genomes. Phylogenetic analyses reveal that the progenitors of ray-finned fish FLERVs co-diversify broadly with their fish hosts, but also suggest that there might have been several ancient viral cross-class transmissions, involving lobe-finned fish, shark, and frog FLERVs. Again, by using the power-law rate-decay model, analyses in this thesis suggest that this major retroviral clade has an ancient Ordovician marine origin, originating together with their jawed vertebrate hosts more than 450 million years ago. This finding implies that the origin of retroviruses as a whole must be in the early Paleozoic Era, if not earlier. The results presented here bridge ancient and recent viral evolution.

In this work we study the existence of weak solutions for compressible Navier-Stokes equations in unbounded time dependent domains. Using the methods introduced in Feireisl E. Dynamics of Viscous Compressible Fluids we extend the results of article Feireisl E. Neustupa J. Stebel J., Convergence of a Brinkman-type penalization for compressible fluid flows, which studies the flow with a "no-slip" boundary condition on bounded domains. Next, we extend results of article Feireisl E. Kreml O. Nečasová Š. Neustupa J. Stebel J., Weak solutions to the barotropic Navier- Stokes system with slip boundary conditions in time dependent domains, which studies flow with compete Navier boundary condition. Finally, we discuss solutions for rotating fluid system. In this case, there are new members in momentum equation, representing the Coriolis and centrifugal force, which cause problems.

Modelos multi-estados têm sido utilizados para descrever o comportamento de unidades amostrais cuja principal resposta é o tempo necessário para a ocorrência de seqüências de eventos. Consideramos um modelo multi-estados markoviano, não homogêneo, que incorpora covariáveis cujos efeitos podem variar ao longo do tempo (efeitos dinâmicos), o que permite a generalização dos modelos usualmente empregados. Resultados assintóticos mostram que procedimentos de estimação baseados no método histograma crivo convergem para um processo gaussiano. A metodologia proposta mostra-se adequada na modelagem de dados reais para comparação de desenvolvimento de recém-nascidos pré-termo com os a termo. Estudos com dados gerados artificialmente confirmam os resultados teóricos obtidos. / Multi-state models have been used to describe the behavior of sample units where the principal response is the time needed for the occurrence of a sequence of events. We consider a non-homogeneous Markovian multi-state model that incorporates covariates with time-dependent coefficient (dynamic effects), generalizing models usually employed. The asymptotic results show that the estimators based on the method of histogram sieves converge to a Gaussian process. The proposed methodology revels adequated for modeling data related to the comparison of developement of preterm infants with term infants. The studies with artificially generated data confirm the asymptotic results.

In spite of the many discoveries made in neuroscience, the mechanism by which memories are formed is still unclear. To better understand how some disorders of the brain arise, it is necessary to improve our knowledge of memory formation in the brain. With the aid of a biological experiment, an artificial neural network is developed to provide insight into how information is stored and recalled. In particular, the bi-conditional association of distinct spatial and non-spatial information is examined using computational techniques. The thesis defines three versions of a computational model based on a combination of feedforward and recurrent neural networks and a biologically-inspired spike time dependent plasticity learning rule. The ability of the computational model to store and recall the bi-conditional object-space association task through reward-modulated plastic synapses is numerically investigated. Further, the network's response to variation of certain parameter values is numerically addressed. A parallel algorithm is introduced to reduce the running time necessary to test the robustness of this artificial neural network. The numerical results produced with this algorithm are then analyzed by a statistical approach, and the network's ability for learning is assessed.

In this research, we investigate and propose new operators to improve Genetic Algorithm’s performance to solve the multi-stop routing problem. In a multi-stop route, a user starts at point x, visits all destinations exactly once, and then return to the same starting point. In this thesis, we are interested in two types of this problem. The first type is when the distance among destinations is fixed. In this case, it is called static traveling salesman problem. The second type is when the cost among destinations is affected by traffic congestion. Thus, the time among destinations changes during the day. In this case, it is called time-dependent traveling salesman problem. This research proposes new improvements on genetic algorithm to solve each of these two optimization problems. First, the Travelling Salesman Problem (TSP) is one of the most important and attractive combinatorial optimization problems. There are many meta-heuristic algorithms that can solve this problem. In this paper, we use a Genetic Algorithm (GA) to solve it. GA uses different operators: selection, crossover, and mutation. Sequential Constructive Crossover (SCX) and Bidirectional Circular Constructive Crossover (BCSCX) are efficient to solve TSP. Here, we propose a modification to these crossovers. The experimental results show that our proposed adjustment is superior to SCX and BCSCX as well as to other conventional crossovers (e.g. Order Crossover (OX), Cycle Crossover (CX), and Partially Mapped Crossover (PMX)) in term of solution quality and convergence speed. Furthermore, the GA solver, that is improved by applying inexpensive local search operators, can produce solutions that have much better quality within reasonable computational time. Second, the Time-Dependent Traveling Salesman Problem (TDTSP) is an interesting problem and has an impact on real-life applications such as a delivery system. In this problem, time among destinations fluctuates during the day due to traffic, weather, accidents, or other events. Thus, it is important to recommend a tour that can save driver’s time and resources. In this research, we propose a Multi-Population Genetic Algorithm (MGA) where each population has different crossovers. We compare the proposed MG against Single-Population Genetic Algorithm (SGA) in terms of tour time solution quality. Our finding is that MGA outperforms SGA. Our method is tested against real-world traffic data [1] where there are 200 different instances with different numbers of destinations. For all tested instances, MGA is superior on average by at least 10% (for instances with size less than 50) and 20% (for instances of size 50) better tour time solution compared to SGA with OX and SGA with PMX operators, and at least 4% better tour time compared toga with SCX operator.

Reliability based life cycle cost analysis is becoming an important consideration for decision-making in relation to bridge design, maintenance and rehabilitation. An optimal solution should ensure reliability during service life while minimizing the life cycle cost. Risk of failure is an important component in whole of life cycle cost for both new and existing structures. Research work presented here aimed to develop a methodology for evaluation of the risk of failure of reinforced concrete bridges to assist in decision making on rehabilitation. Methodology proposed here combines fault tree analysis and probabilistic time-dependent reliability analysis to achieve qualitative and quantitative assessment of the risk of failure. Various uncertainties are considered including the degradation of resistance due to initiation of a particular distress mechanism, increasing load effects, changes in resistance as a result of rehabilitation, environmental variables, material properties and model errors. It was shown that the proposed methodology has the ability to provide users two alternative approaches for qualitative or quantitative assessment of the risk of failure depending on availability of detailed data. This work will assist the managers of bridge infrastructures in making decisions in relation to optimization of rehabilitation options for aging bridges.

In this paper, the numerical simulation for nano-indentation is performed to measure time-dependent behavior of polymeric films. The possibility to extract the relaxed shear modulus of the polymer is evaluated using a rigid ball indenter. The viscoelastic behavior of the polymer was represented by the standard model. The effects of Poisson’s ratio are also discussed. / Singapore-MIT Alliance (SMA)