Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 30
Encontrada(s) 1 página(s)

BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Fast One-dimensional Finite Element Approximation of Geophysical Measurements

  • Shahriari, M.
There exist a wide variety of geophysical prospection methods. In this work, we focus on resistivity methods. We categorize these resistivity prospection methods according to their acquisition location as (a) on the surface, such as the ones obtained using Controlled Source Electromagnetics (CSEM) and magnetotelluric, and (b) in the borehole, such as the ones obtained using Logging-While-Drilling (LWD) devices. LWD devices are useful both for reservoir characterization and geosteering purposes, which is the act of adjusting the tool direction to travel within a specific zone. When inverting LWD resistivity measurements, it is a common practice to consider a one-dimensional (1D) layered media to reduce the problem dimensionality using a Hankel transform. Using orthogonality of Bessel functions, we arrive at a system of Ordinary Differential Equations (ODEs); one system of ODEs per Hankel mode. The dimensionality of the resulting problem is referred to as 1.5D since the computational cost to resolve it is in between that needed to solve a 1D problem and a 2D problem. When material properties (namely, resistivity, permittivity, and magnetic permeability) are piecewise-constant, we can solve the resulting ODEs either (a) analytically, which leads to a so-called semi-analytic method after performing a numerical inverse Hankel transform or (b) numerically. Semi-analytic methods are faster, but they also have important limitations, for example, (a) the analytical solution can only account for piecewise constant material properties, and other resistivity distributions cannot be solved analytically, which prevents to accurately model, for example, an Oil-Water Transition (OWT) zone when fluids are considered to be immiscible; (b) a specific set of cumbersome formulas has to be derived for each physical process (e.g., electromagnetism, elasticity, etc.), anisotropy type, etc.; (c) analytical derivatives of specific models (e.g., cross-bedded formations, or derivatives with respect to the bed boundary positions) are often difficult to obtain and have not been published to the best of our knowledge. In view of the above limitations, we propose to solve our forward problems using a numerical solver. A traditional Finite Element Method (FEM) is slow, which makes it unfeasible for our application. To achieve high performance, we developed a multiscale FEM that pre-computes a set of optimal local basis functions that are used at all logging positions. The resulting method is slow when compared to a semi-analytic approach for a single logging position, but it becomes highly competitive for a large number of logging positions, as needed for LWD geosteering applications. Moreover, we can compute the derivatives using an adjoint state method at almost zero additional cost in time. We describe an adjoint-based formulation for computing the derivatives of the electromagnetic fields with respect to the bed boundary positions. The key idea to obtain this adjoint-based formulation is to separate the tangential and normal components of the field, and treat them differently. We then apply this method to a 1.5D borehole resistivity problem. Moreover, we compute the adjoint-state formulation to compute the derivative of the magnetic field with respect to the resistivity value of each layer. We verify the accuracy of our formulations via synthetic examples. When simulating borehole resistivity measurements in a reservoir, it is common to consider an Oil-Water Contact (OWC) planar interface. However, this consideration can lead to an unrealistic model since, in the presence of capillary pressure, the mix of two immiscible fluids (oil and water) often appears as an OWT zone. These transition zones may be large in the vertical direction (20 meters or above), and in context of geosteering, an efficient method to simulate an OWT zone can maximize the production of an oil reservoir. In this work, we prove that by using our proposed 1.5D numerical method, we can easily consider arbitrary resistivity distributions in the vertical direction, as it occurs in an OWT zone. Numerical results on synthetic examples demonstrate significant differences between the results recorded by a geosteering device when considering a realistic OWT zone vs. an OWC sharp interface. As an additional piece of work of this Ph.D. Dissertation, we explore the possibility of using a Deep Neural Network (DNN) to perform a rapid inversion of borehole resistivity measurements. Herein, we build a DNN that approximates the following inverse problem: given a set of borehole resistivity measurements, the DNN is designed to deliver a physically meaningful and data-consistent piecewise one-dimensional layered model of the surrounding subsurface. Once the DNN is built, the actual inversion of the field measurements is efficiently performed in real time. We illustrate the performance of a DNN designed to invert LWD measurements acquired on high-angle wells via synthetic examples.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Computational predictive modeling of integrated cerebral metabolism, electrophysiology and hemodynamics

  • Capo, G.
Understanding the energetic requirement of brain cells during resting state and during high neuronal activity is a very active research area where mathematical models have contributed significantly by providing a context for the interpretation of the experimental results. In this thesis, we present three new computational predictive mathematical models to elucidate several dynamics in the brain, comprising electrophysiological activity, cellular metabolism and hemodynamic response. Many computational challenges had to be addressed, mostly due to the very different characteristic times at which the electrical, metabolic and hemodynamic events occur. The first part of the thesis proposes a novel predictive mathematical electro-metabolic model connecting the electrophysiological activity and the metabolism through a double feedback mechanism based on energy demand and production. This model sheds light on the role of the glial potassium cleaning in brain energy metabolism by integrating a four compartment metabolic model with one describing in details the electrical activity. The results of computed experiments performed with this model for different protocols, namely awake resting state, transitions between resting state and neuronal activation and ischemic episodes are in agreement with experimental observations. In the second part of the thesis, the electro-metabolic model is expanded to comprise the brain hemodynamic response. This is attained through a triple feedback mechanism between the electrophysiology, metabolism and a three compartment hemodynamic model tracking the changes of cerebral blood flow and cerebral blood volume through arteries, capillaries and veins. During neuronal activation, the increase in extracellular potassium concentration triggers an increase in the cerebral blood flow and concurrently vasodilation, ensuring the supply of nutrients necessary for the metabolic response to sustain the increased energy demand. The ensuing hemo-electro-metabolic model provides a better insight on the transitions between resting state and neuronal activation. In the third and last part of the thesis, we propose a variant of the electro-metabolic model that adequately describes the changes in the brain in connection with cortical spreading depression (CSD) waves. In addition the dynamics of sodium and potassium, the new model accounts for chloride dynamics, the glutamate-glutamine cycle, as well as neuronal swelling accompanied by shrinkage of extracellular space. As illustrated with computed experiments, with this model it is possible to follow simultaneously the changes in ionic homeostasis, the alterations in the volumes of the cellular compartments and of the extracellular space, and large modifications in brain metabolism during cortical spreading depression waves. The model predictions, in agreement with findings reported in the experimental literature, show a large decrease in glucose and oxygen concentration and a significant increase in lactate concentration during the passing of cortical spreading depression waves., SVP-2014-068725
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Reduced description method in the kinetic theory of Brownian motion with active fluctuations

  • Sliusarenko, O.
  • Sliusarenko, Y.
We develop a microscopic approach to the kinetic theory of many-particle systems with dissipative and potential interactions in presence of active fluctuations. The approach is based on a generalization of Bogolyubov–Peletminskii reduced description method applied to the systems of many active particles. It is shown that the microscopic approach developed allows to construct the kinetic theory of two- and three-dimensional systems of active particles in presence of non-linear friction (dissipative interaction) and an external random field with active fluctuations. The kinetic equations for these systems in case of a weak interaction between the particles (both potential and dissipative ones) and low-intensity active fluctuations are obtained. We demonstrate particular cases in which the derived kinetic equations have solutions that match the results known in the literature. In addition, analysis of particular solutions showed that in the case of a friction force linearly dependent on the speed of structural units, the manifestation of self-propelling properties in the process of evolution of an active medium may be due to the local nature of active fluctuations.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Asymptotic behaviour of some nonlocal equations in mathematical biology and kinetic theory

  • Havva, Y.
We study the long-time behaviour of solutions to some partial differential equations arising in modeling of several biological and physical phenomena. In this work, the type of the equations we consider is mainly nonlocal, in the sense that they involve integral operators. %In order to check if the equation is valid at one point, we need an information about the region far from that point. Moreover, the equations we consider describe the time evolution of either some populations structured by several traits like age, elapsed-time and size or the distribution of the dynamical states of a single particle, depending on time, space and velocity. In the latter case, they are called kinetic equations. We are interested in showing quantitatively the asynchronous behaviour of interacting neuron populations which are composed of large and fully connected networks. Neurons undergo a charging period followed by a sudden discharge in the form of firing a spike. We consider two nonlinear models structured by the time elapsed since the last discharge and nonlinearity comes from the dependence of firing rate on the total neural activity at a time. In the second model, there is an addition of a fragmentation term to include the effect of the past activity of neurons by displaying adaptation and fatigue. With this addition, the equation shares many common properties with another class of integro-partial differential equations called the growth-fragmentation equation. This is the second type of equation we look at the convergence rate to a universal profile in a quantitative way. The growth-fragmentation equation describes a system of growing and dividing particles which may be used as a model for many processes in ecology, neuroscience, telecommunications and cell biology. We consider two types of fragmentation processes, namely mitosis and constant fragmentation and include nonconservative cases where eigenelements cannot be computed explicitly. We present quantitative exponential convergence speeds in the weighted total variation norm. Furthermore, we also study hypocoercivity of some space inhomogeneous linear kinetic equations including linear relaxation Boltzmann(linear BGK) and linear Boltzmann equations which are posed either on the torus or on the whole space with a confining potential. We prove exponential convergence in the torus or on the whole space with a potential growing quadratically at infinity. Moreover, for the weaker confining potentials (subquadratic) we present subgeometric convergence rates quantitatively. The physiologically structured population models and the space inhomogeneous linear kinetic equations we deal with in this work are well-studied from various aspects in the already-existing literature. We provide the references later. What differs from the past plentiful studies on the asymptotic behaviour of these equations is the techniques we use here. We consider a probabilistic approach which is first developed for studying ergodic properties of discrete-time Markov processes. The method is due to Doeblin and Harris; based on establishing a combination of a minorisation (irreducibility) and a geometric drift (Lyapunov) conditions for a Markovian process. This method gives a quantitative convergence speed and existence of a unique steady state even without having to calculate it explicitly. Application of Harris's Theorem into the aforementioned partial differential equations to study the long-time behaviour of solutions is the core of this thesis., la Caixa Foundation
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Stochastic processes for anomalous diffusion

  • Molina-Garcia, D.
Anomalous diffusion is a diffusion process which Mean Square Displacement (MSD) is not a linear funtion of time, what is known as normal diffusion. When the relation is faster than linear, it is called superdiffusion and when it is slower, subdiffusion. Physically, the MSD is a measure of the deviation of the position of the particles over time. It can be thought as the amount of space the particles have explored in the system. On the one hand, anomalous diffusion has been shown to appear extensively in nature and, consecuently, scientists have developed several and different models that can effectively reproduce it. However, the underlying physics of a plethora of experiments is still not well-understood. This is the case of the motion of mRNA molecules inside living E. coli cells, where stochastic processes as the continuous-time random walk can explain non-ergodicity but an alternative like the fractional Brownian motion is needed to reproduce p-variation. We find that a family of stochastic processes known as generalised grey Brownian motion can fit correctly that observation. In addition, we also use an explicit time-dependent factor to model non-stationarity, usually referred to aging in the specialized literature. In agreement with our computational results, we were able to obtain analytical expressions for a list of observables too, including temporal and ensemble-average MSD, Ergodicity Breaking parameter, one-point one-time probability density function and p-variation. On the other hand, many experiments show a characteristic crossover from anomalous to normal diffusion. For example, it happens in viscoelastic systems such as the motion of lipid molecules in lipid bilayer membranes. We studied the motion of particles driven by tempered fractional Gaussian noise which power-law correlations present a cutoff at some mesoscopic time scale. Deriving analytical expressions of the MSD for the overdamped Langevin equation and the fractional Langevin equation, we find that when the truncation is strongly done, the mentioned crossover appears. When the truncation is done by a weak power-law truncation, we also got a different crossover behaviour from faster to slower superdiffusion and from slower to faster subdiffusion. We were also interested in another process defined through a tempering directly done in the fractional Brownion motion definition. It is known as tempered fractional Brownian motion and, surprisingly, it does not arrive to the same behaviours than our models. Instead, at long times it exhibits localization as Ornstein-Uhlenbeck. We compare both of them. When the derivative of this tempered fractional Brownian motion is used in the fractional Langevin equation, it leads to ballistic diffusion at long times. In the last part, we were interested in the state of the art in spatial neutral theory of Ecology, which problems sound familiar to experts in Statistical Physics. In fact, ecosystems display a complex spatial organization that ecologists have tried to characterize by observing patterns of biodiversity at different spatial scales. Linking those measures with the causes that originate them is arguably the central problem in Ecology. Ecological neutral theory, which underscores the role of stochastic demographic fluctuations and neglects deterministic effects stemming from fitness differences, has predicted the empirical patterns in communities of competing species. We study non-trivial scaling laws arising at the critical dimension (2D) of spatial neutral models, which are poorly understood. We conclude by discussing models that support non-neutral features.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Advanced Mathematical Modelling of Pancreatic β-Cells

  • Marinelli, I.
Insulin-secreting pancreatic $\beta$-cells are responsible for maintaining the whole body glucose homeostasis. Dysfunction or loss of $\beta$-cell mass results in impaired insulin secretion and, in some cases, diabetes. Many of the factors that influence $\beta$-cell function or insulin exocytosis, however, are not fully understood. To support the investigation, mathematical models have been developed and used to design experiments. In this dissertation, we present the Integrated Oscillator Model (IOM) that is one of the mathematical models used for the investigation of the mechanism behind the bursting activity that underlies intracellular Ca$^{2+}$ oscillations and pulsatile insulin secretion. The IOM describes the interaction of the cellular electrical activity and intracellular Ca$^{2+}$ with glucose metabolism via numerous feedforward and feedback pathways. These interactions, in turn, produce metabolic oscillations with a sawtooth or pulsatile time course, reflecting different oscillation mechanisms. We determine conditions favorable to each type of oscillations, and show that the model accounts for key experimental findings of $\beta$-cell activity. We propose several extensions of the model to include all the main elements involved in the insulin secretion. The latest and most sophisticated model describes the complex metabolism in the mitochondria and the several biological processes in the insulin exocytosis cascade. The model, also, captures the changes in the $\beta$-cell activity and the resulting amount of secreted insulin in response to different concentrations of glucose in the blood. The model predictions, in agreement with findings reported in the experimental literature, show an increase of insulin secretion when the glucose level is high and a basal-low insulin concentration when the glucose level decreases. Finally, we use the new model to simulate the interaction among $\beta$-cells (through gap junction) within the same islet. The simulations show that the electrical coupling is sufficient to synchronize the $\beta$-cells within an islet. We also show that the amplitude of the oscillations in the insulin secretion rate is bigger when the $\beta$-cells synchronize. This suggests a more efficient secretion of insulin in the bloodstream when the cells burst in unison, as it has been observed experimentally.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

K-means for massive data

  • Capo, M.
The $K$-means algorithm is undoubtedly one of the most popular clustering analysis techniques, due to its easiness in the implementation, straightforward parallelizability and competitive computational complexity, when compared to more sophisticated clustering alternatives. However, the progressive growth of the amount of data that needs to be analyzed, in a wide variety of scientific fields, represents a significant challenge for the $K$-means algorithm, since its time complexity is dominated by the number of distance computations, which is linear with respect to both the number of instances, $n$, and dimensionality of the problem, $d$. This fact hinders its scalability on such massive data sets. Another major drawback of the $K$-means algorithm corresponds to its high dependence on the initial conditions, which not only may affect the quality of the obtained solution, but may also have a major impact on its computational load, as for instance, a poor initialization could lead to an exponential running time in the worst case scenario. In this dissertation, we tackle all these difficulties. Initially, we propose an approximation to the $K$-means problem, the Recursive Partition-based $K$-means algorithm (RP$K$M). This approach consists of recursively applying a weighted version of $K$-means algorithm over a sequence of spatial-based partitions of the data set, for which each cell of the partition is represented by the center of mass of the points that lie on it. From one iteration to the next, a more refined partition is constructed and the process is repeated using the optimal set of centroids, obtained at the previous iteration, as initialization. From a practical standpoint, such a process reduces the computational load of $K$-means algorithm as the number of representatives, at each iteration, is commonly much smaller than the number of instances of the data set. On the other hand, both phases of the algorithm are embarrasingly parallel. From a theoretical standpoint, and in spite of the selected partition strategy, one can guarantee the non-repetition of the clusterings generated at each RP$K$M iteration, which ultimately implies the reduction of the total amount of $K$-means algorithm iterations, as well as leading, in most of the cases, to a monotone decrease of the overall error function. Afterwards, we present a RP$K$M-type approach, the Boundary Weighted $K$-means algorithm (BW$K$M). For this technique, the data set partition is based on an adaptative mesh, that adjusts the size of each grid cell to maximize the chances of each cell having only instances of the same cluster. The goal is to focus most of the computational resources on those regions where it is harder to determine the correct cluster assignment of the original instances (which is the main source of error for our approximation). For such a construction, it can be proved that if all the cells of a spatial partition are well assigned (have instances of the same cluster) at the end of a BW$K$M step, then the obtained clustering is actually a fixed point of the $K$-means algorithm over the entire data set. Furthermore, if, for a certain step of BW$K$M, this property can be verified at consecutive weighted Lloyd’s iterations, then the error of our approximation also decreases monotonically. From a practical stand point, BW$K$M was compared to the state-of-the-art: $K$-means++, Forgy $K$-means, Markov Chain Monte Carlo $K$-means and Minibatch $K$-means. The obtained results show that BW$K$M commonly converged to solutions, with a relative error of under $1\%$ with respect to the considered methods, while using a much smaller amount of distance computations (up to 7 orders of magnitude lower). Even when the computational cost of BW$K$M is linear with respect to the dimensionality, its error quality guarantees are mainly related to the diagonal length of the grid cells, meaning that, as we increase the dimensionality of the problem, it will be harder for BW$K$M to have such a competitive performance. Taking this into consideration, we developed a fully-parellelizable feature selection technique intended for the $K$-means algorithm, the Univariate $K$-means relevance for feature selection algorithm ($K$MR). This approach consists of applying any heuristic for the $K$-means problem over multiple subsets of dimensions (each of which is bounded by a predefined constant, $m\ll d$) and using the obtained clusterings to upper-bound the increase in the $K$-means error when deleting a given feature. We then select the features with the $m$ largest error increases. Not only can each step of $K$MR be simply parallelized, but its computational cost is dominated by that of the selected heuristic (on $m$ dimensions), which makes it a suitable dimensionality reduction alternative for BW$K$M on large data sets. Besides providing a theoretical bound for the obtained solution, via $K$MR, with respect the optimal $K$-means clustering, we analyze its performance in comparison to well-known feature selection and feature extraction techniques. Such an analysis shows $K$MR to consistently obtain results with lower $K$-means error than all the considered feature selection techniques: Laplacian scores, maximum variance and random selection, while also requiring similar or lower computational times than these approaches. On the other hand, $K$MR, when compared to feature extraction techniques, such as Random Projections, also shows a noticeable improvement in both error and computational time. As a response to the high dependency of $K$-means algorithm to its initialization, we finally introduce a cheap and yet effective Split-Merge step that can be used to re-start the $K$-means algorithm after reaching a fixed point, Split-Merge $K$-means (SM$K$M). Under some settings, one can show that this approach reduces the error of the given fixed point without requiring any further iteration of the $K$-means algorithm. Moreover, experimental results show that this strategy is able to generate approximations with an associated error that is hard to reach for different multi-start methods, such as multi-start Forgy $K$-means, $K$-means++ and Hartigan $K$-means. In particular, SM$K$M consistently generated the local minima with the lowest $K$-means error, reducing, on average, over 1 and 2 orders of magnitude of relative error with respect to $K$-means++ and Hartigan $K$-means and Forgy $K$-means, respectively. We have observed that SM$K$M improves the previously commented methods in terms of the number of distances computed and the error of the obtained solutions.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Patient-specific modelling of cortical spreading depression applied to migraine studies

  • Kroos, J.M.
Migraine is a common neurological disorder and one-third of migraine patients suffer from migraine aura, a perceptual disturbance preceding the typically unilateral headache. Cortical spreading depression (CSD), a depolarisation wave that originates in the visual cortex and propagates across the cortex to the peripheral areas, has been suggested as a correlate of visual aura by several studies. The complex and highly individual-specific characteristics of the brain cortex suggest that the geometry might have a significant impact on CSD propagation. In this thesis, we combine two existing models, a detailed neurological model for the electrophysiological component of CSD and a reaction-diffusion model accounting for the potassium diffusion, the driving force of CSD propagation. In the process, we integrate two aspects of CSD that occur at different time scales: the electrophysiological dynamics features a temporal scale in the order of milliseconds, while the extracellular potassium dynamics that triggers CSD propagation features is on the scale of minutes. As a result we obtain a multi-scale PDE-ODE model. In addition, we incorporate patient-specific data in the CSD model: (i) a patient-specific brain geometry obtained from magnetic resonance imaging, and (ii) personalised conductivity tensors derived from diffusion tensor imaging data. To study the role of the geometry in CSD propagation, we define geometric and CSD-dependent quantities of interest (QoI) that we evaluate in two case studies. Even though the geometry does not seem to have a major impact on the CSD propagation, some QoI are promising candidates to aid in the classification of healthy individuals and migraine patients. Finally, to account for the lack of experimental data for validation and selection of the model parameters, we apply different techniques of uncertainty quantification to the CSD model and analyse the impact of various parameter choices on the model outcome.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Snapshots location for Reduced Order Models: An approach based on Proper Orthogonal Decomposition and Mesh Adaptivity Techniques

  • Bidaguren, I.
Solving Partial Differential Equations (PDE), is a key issue in science and engineering since they are widely used in many real problem modelling including fluid mechanics, acoustics, heat and mass transfer ...etc. Exact solutions for PDEs and Ordinary Differential Equations (ODE) can be obtained in a very few and simplified cases only, numerical approximations are used instead (i.e. Finite Volume Method (FEM), Finite Volume Method (FVM)...). However, and despite the availability of important supercomputing facilities, due to the huge number of degrees of freedom these methods have, they may still suffer from cost-effectiveness performance. They are mainly two contexts where one need to solve PDEs with a lower computational cost: Real-time context and Many-query context. Examples for the former are: parameter-estimation, control, flying simulator..., and for the latter: optimization, multi model/scale simulation... To reduce significantly simulation time (often on the expense of accuracy) Reduced Order Modelling (ROM) techniques are introduced. The main idea is to reduce the initial solving space dimension (as in finite elements method) to a subspace with a significantly reduced dimension and then solve for the projected solutions. Proper Orthogonal Decomposition (POD) is one of the more used ROM strategies. The presented work focuses on this technique, which has two challenging steps: (i) the snapshot location and (ii) the error estimate on the parameter space that drive the process to search new snapshot location. As a consequence of these two steps, POD applied to PDEs is considered as belonging to the well-known Greedy Algorithm family. This thesis brings a mesh adaptivity approach as the process to find the new parameter space locations. This process will be driven by a new error estimate based on Leave One Out Cross Validation (LOOCV) technique. We could say that this error estimate is universal, in the sense that it is not problem dependent. In addition, it is well known POD lack of accuracy when dealing with PDEs which solution contains shocks. Here, a new interpolation approach improves this, for shocked solutions. Finally, we present that the POD reduced basis is optimal just in average, and a new local basis is presented (Sorted Gram-Schmidt (SGS)) to be coupled with POD one. The criteria to decide which basis is better to be used for each new parameter value is defined as well. The whole proposed strategy efficiency is validated against a mathematical (exact) solutions of an incompressible, steady state flow equations, and on CFD solutions of an inviscid flow around a NACA0012 airfoil.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

A General Framework for Prediction in Generalized Additive Models

  • Carballo, A.
Smoothing techniques have become one of the most popular modelling approaches in the unidimensional and multidimensional setting. However, out-of-sample prediction in the context of smoothing models is still an open problem that can significantly widen the use of these models in many areas of knowledge. The objective of this thesis is to propose a general framework for prediction in penalized regression, particularly in the P-splines context. To that end, Chapter 1 includes a review of the different proposals available in the litera- ture, and results useful and necessary along the document, the formulation of a P-spline model and its reparameterization as a mixed model. In Chapter 2, we generalize the approach given by Currie et al. (2004) to predict with any regression basis and quadratic penalty. For the particular case of penalties based on differences between adjacent coefficients, we reparameterize the extended P-spline model as a mixed model and we prove that the fit remains the same as the result we obtain only fitting the data and show the crucial role of the penalty order, since it determines the shape of the prediction. Moreover, we adapt available methods in contexts such as mixed models (Gilmour et al. 2004) or global optimization (Sacks et al. 1989) to predict in the context of penalized regression and prove their equivalence for the particular case of P- splines. An extensive section of examples illustrates the application of the methodology. We use three real datasets with particular characteristics: one of them on aboveground biomass allow us to show that prediction can also be performed to the left of the data; other of them, on monthly sulphur dioxide levels, illustrates how prediction can take into account the temporal trends and seasonal effects by using the smooth modulation model based on P-splines suggested by Eilers et al. (2008); and other, on yearly sea level, shows that prediction can also be carried out in the case of correlated errors. We also introduce the concept of “memory of a P-spline” as a tool to know how much of the known information we use to predict new values. In the third chapter, we propose a general framework for prediction in multidimensional smoothing, we extend the proposal of Currie et al. (2004) to predict when more than one covariate is extended. The extension of the prediction method to the multidimensional case is not straightforward in the sense that, in this context, the fit changes when the fit and the prediction are carried out simultaneously. To overcome this problem we propose an easy but elegant solution, based on Lagrange multipliers. The first part of the chapter is dedicated to show how out-of-sample predictions can be carried out in the context of multidimensional P-splines and the properties satisfied, under certain conditions, by the coefficients that determine the prediction. We also propose the use of restrictions to maintain the fit, and in general, to incorporate any known information about the prediction. The second part of the chapter is dedicated to extend the methodology to the smooth mixed model framework. It is known that when a P-spline model is reparameterized as a mixed model, the structure of the coefficients is lost, that is, they are not ordered according to the position of the knots. This fact is not relevant when we fit the data, but if we predict and impose restrictions over the coefficients, we need to differentiate between the coefficients that determine the fit and the coefficients that determine the prediction. In order to do that, we define a particular transformation matrix that preserves the original model matrices. The prediction method and the use of restrictions is illustrated with one real data example on log mortality rates of US male population. We show how to solve the crossover problem of adjacent ages when mortality tables are forecasted and compare the results with the well-known method developed in Delwarde et al. (2007). The research in Chapter 4, is motivated by the need to extend the prediction methodol- ogy in the multidimensional case to more flexible models, the so-called Smooth-ANOVA models, which allow us to include interaction terms that can be decomposed as a sum of several smooth functions. The construction of these models through B-splines basis suffer from identifiability problems. There are several alternatives to solve this problem, here we follow Lee and Durb ́an (2011) and reparameterize them as mixed models. The first two sections of the chapter are dedicated to introduce the Smooth-ANOVA models and to show how out-of-sample prediction can be carried out in these models. We illustrate the prediction with Smooth-ANOVA models reanalyzing the dataset on aboveground biomass. Now, the Smooth-ANOVA model allows us to represent the smooth function as the sum of a smooth function for the height, a smooth function for the diameter of a tree, and a smooth term for the height-diameter interaction. At the end of this chapter, we provide a simulation study in order to evaluate the accuracy of the 2D interaction P-spline models and Smooth-ANOVA models, with and without imposing invariance of the fit. From the results of the simulation study, we conclude that in most situations the constrained S-ANOVA model behaves better in the fit and out-of-sample predictions,however, results depend on the simulation scenario and on the number of dimensions in which the prediction is carried out (one or both dimensions). In the fifth chapter we generalize the developed methodology for generalized linear mod- els (GLMs) in the context of P-splines (P-GLMs) and mixed models (P-GLMMs). In both frameworks, the coefficients and parameters estimation procedures involve nonlinear equations. To solve them iterative algorithms based on the Newton-Raphson methods are used, regardless of the estimation criterion used (for instance, in the GLMMs context we can maximize the residual maximum likelihood (REML) or an approximate REML (based on Laplace approximation)). These iterative algorithms are based on a working normal theory model or a set of pseudodata and weights. Based on this idea, we ex- tend the Penalized Quasilikelihood method (PQL) to fit and predict simultaneously in the context of GLMM. We highlight that, in the context of mixed models (even in the univariate case), to maintain the fit a transformation that preserves the original model matrices has to be used, since different transformations deal with different working vec- tors and therefore with different solutions. We also show how restrictions can be imposed in P-GLMs and P-GLMMs models. To illustrate the procedures we use a real dataset to predict deaths due to respiratory disease through 2D interaction P-splines and S-ANOVA models (both with and without the restriction the fit has to be maintained). Finally, Chapter 6 is devoted to summarize the main conclusions and pose a list of future lines of work.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Vortex Filament Equation for some Regular Polygonal Curves

  • Kumar, S.
One of the most interesting phenomena in fluid literature is the occurrence and evolution of vortex filaments. Some of their examples in the real world are smoke rings, whirlpools, and tornadoes. For an ideal fluid, there have been several models and governing equations to describe this evolution; however, due to its simplicity and geometric properties, the vortex filament equation (VFE) has recently gained attention. As an approximation of the dynamics of a vortex filament, the equation first appeared in the work of Da Rios at the beginning of the twentieth century. This model is usually known as the local induction approximation (LIA). In this work, we examine the evolution of VFE for regular polygonal curves both from a numerical and theoretical point of view in the Euclidean as well as hyperbolic geometry. In the first part of the thesis, we observe the evolution of the Vortex Filament equation taking $M$-sided regular polygons with nonzero torsion as initial data in the Euclidean space. Using algebraic techniques, backed by numerical simulations, we show that the solutions are polygons at rational times, as in the zero-torsion case. However, unlike in that case, the evolution is not periodic in time; moreover, the multifractal trajectory of the point X(0, t) is not planar and appears to be a helix for large times. These new solutions of VFE can be used to illustrate numerically that the smooth solutions of VFE given by helices and straight lines share the same instability as the one already established for circles. This is accomplished by showing the existence of variants of the so-called Riemann’s non-differentiable function that are as close to smooth curves as desired when measured in the right topology. This topology is motivated by some recent results on the well-posedness of VFE, which prove that the self-similar solutions of VFE have finite renormalized energy. In the rest of the work, we delve into the hyperbolic setting and examine the evolution of VFE for a planar $l$-polygon, i.e., a regular planar polygon in the Minkowski 3-space $\mathbb{R}^{1,2}$. Unlike in the Euclidean case, a planar $l$-polygon is open which makes the problem more challenging from a numerical point of view. After trying several numerical methods, we conclude that a finite-difference discretization in space combined with an explicit Runge--Kutta method in time, gives the best numerical results both in terms of efficiency and accuracy. On the other hand, using theoretical arguments, we recover the evolution algebraically, and thus, we show the agreement between the two approaches. During the numerical evolution, it has been observed that the trajectory of a corner is multifractal and as the parameter $l$ goes to zero, it converges to the Riemann’s non-differentiable function. Furthermore, as in the Euclidean case, we provide strong numerical evidence to show that at infinitesimal times, the evolution of VFE for a planar $l$-polygon as an initial datum can be described as a superposition of several one-corner initial data. As a consequence, not only we can compute the speed of the center of mass of the planar $l$-polygon theoretically, the relationship also reveals important properties of the trajectory of its corners which we compare with its equivalent in the Euclidean case. Finally, a nonzero torsion in the hyperbolic case, yields two different kinds of helical polygonal curves, however, with the numerical and theoretical techniques developed so far, we are able to address them as well. This remains part of the future work of the thesis.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

A geometric and physical study of Riemann's non-differentiable function

  • Eceizabarrena, D.
Riemann's non-differentiable function is a classic example of a continuous but almost nowhere differentiable function, whose analytic regularity has been widely studied since it was proposed in the second half of the 19th century. But recently, strong evidence has been found that one of its generalisation to the complex plane can be regarded as the trajectory of a particle in the context of the evolution of vortex filaments. It can, thus, be given a physical and geometric interpretation, and many questions arise in these settings accordingly. It is the purpose of this dissertation to describe, study and prove geometrically and physically motivated properties of Riemann's non-differentiable function. In this direction, a geometric analysis of concepts such as the Hausdorff dimension, geometric differentiability and tangents will be carried out, and the relationship with physical phenomena such as the Talbot effect, turbulence, intermittency and multifractality will be explained., Ministerio de Educación, Cultura y Deporte - FPU15/03078
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

The Schrödinger equation and Uncertainty Principles

  • Aguirre, M.
The main task of this thesis is the analysis of the initial data u0 of Schrödinger’s initial value problem in order to determine certain properties of its dynamical evolution. First we consider the elliptic Schrödinger problem in its perturbative form. The idea is to find lower bounds for the solution giving conditions at time t = 0 together with a size condition on the potential. After analyzing the elliptic case we give a similar result for the hyperbolic Schrödinger operator. Next we focus on the free particle case, this is, the case where no potential is involved. The goal here is to quantify the L2 norm of the solution in a space-time cylinder. Following the same idea as before we want to find conditions at time t = 0 to ensure this. To carry out this task we define the Σδ space where δ is a parameter on the interval (0,1]. We see that if u0 belongs in this space then so does its evolution in time and use this fact to give lower bounds for the L2 norm of the solution. For δ = 1 we give a different approach and make use of the Virial Theorem. We will see that this case has particular properties. Finally we study dynamical uncertainty principles derived from the previous study. The key point will be to write the solution as u = ρeiθ, where ρ and θ are real functions. Thus we give uncertainty principles in terms of these functions and find explicit expressions for them so that u becomes a minimizer of the problem.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Generalized Poincaré-Sobolev inequalities

  • Martínez-Perales, J.
Poincaré-Sobolev inequalities are very powerful tools in mathematical analysis which have been extensively used for the study of differential equations and their validity is intimately related with the geometry of the underlying space. In particular, and since their applicability as part of the Moser iteration method, their weighted counterparts are of interest for applications. The goal of this dissertation is to present a self-contained study of Poincaré-Sobolev inequalities, weights and the combination of both under the framework of the abstract theory of generalized Poincaré-Sobolev inequalities. To this end, the basic aspects on the theory of Poincaré-Sobolev inequalities and the theory of Muckenhoupt weights is presented. In relation with these, the class of functions with bounded mean oscillations is studied, together with a new characterization of it through some boundedness properties of commutators of fractional integrals. A unified study of classical and fractional weighted Poincaré-Sobolev inequalities, as well as a study of Muckenhoupt weights in relation with functions with bounded mean oscillations is carried out by using new self-improving techniques., BERC 2014-2017, SEV-2013-0323
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Reflection maps

  • Peñafort, G.
Given a reflection group G acting on a complex vector space V , a reflection map is the composition of an embedding X → V with the quotient map V → Cp of G. We show how these maps, which can highly singular, may be studied in terms of the group action. We give obstructions to A-stability and A-finiteness of reflection maps and produce, in the unobstructed cases, infinite families of finitely determined map-germs of any corank. We relate these maps to conjectures of Lê, Mond and Ruas.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Mathematical models for glioma growth and migration inside the brain

  • Conte, M.
Gliomas are the most prevalent, aggressive, and invasive subtype of primary brain tumors, characterized by rapid cell proliferation and great infiltration capacity. De- spite the advances of clinical research, these tumors are often resistant to treatment, the median survival ranges between 9 and 12 months, and recurrence is the main cause of mortality. Glioma migration and invasion into the brain tissue is a complex phenomenon and little is still known about the underlying mechanisms that lead to tumor progression. In this thesis, we propose several mathematical models studying various aspects of glioma progression in relation to the microscopic and macroscopic scales charac- terizing this process. Exploiting the inherently multiscale nature of glioma evolution allows to define models based on dynamical systems, kinetic equations, and macro- scopic PDEs with different roles depending on the considered phenomena. The in- tegration of biological and clinical data with the mathematical models is one of the key objectives of this thesis. The experimental data at hand are obtained from mag- netic resonance and diffusion tensor images of the human brain and from in-vivo im- munofluorescence analysis of protein distributions in Drosophila, a reliable model for the study of glioblastoma dynamics. We analyze the anisotropic characteristics of the brain tissue, using the diffusion tensor data, and the influence of the fiber structures on tumor cell dynamics. We show how the fiber network directs cell migration along preferential paths, reproducing the branched and heterogeneous patterns typical of glioma evolution, and how multi- modal treatments can reduce this behavior. We study the interdependency of microenvironmental acidity and vasculature in tumor angiogenesis, defining a model capable of reproducing their influence on the emergence of phenotypic heterogeneity and hypoxia-related features (like necrosis) typical of glioma progression. This study enables the testing of a necrosis-based tumor grading and the investigation of multi-modal therapies with anti-angiogenic effects. We investigated the role of cell protrusions from a non-local perspective. We ex- plore their influence on the contact guidance phenomenon and on the emergence of collaborative or competitive effects between two cues driving cell velocity changes. Using the experimental analysis of protein distributions, we evaluate cell protru- sion relationship with integrins and proteases as leading mechanisms of glioblastoma progression. We show how the biochemical and biomechanical interactions of these agents result in the emergence of tumor propagation fronts, which can feature a dy- namical and heterogenous evolution in relation to environmental changes., European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 713673. ”la Caixa” Foundation (ID 100010434), with fellowship code LCF/BQ/IN17/11620056.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Algorithms for Large Orienteering Problems

  • Kobeaga, G.
In this thesis, we have developed algorithms to solve large-scale Orienteering Problems. The Orienteering Problem is a combinatorial optimization problem were given a weighted complete graph with vertex profits and a maximum distance constraint, the goal is to find the simple cycle which maximizes the sum of the profits of the visited vertices. To solve the Orienteering Problem, we have developed an evolutionary algorithm and an Branch-and-Cut algorithm. One of the key characteristics of the evolutionary algorithm is to work with unfeasible solutions. From the point of view of genetic operators, the main contribution has been the development of the Edge Recombination Crossover for the Orienteering Problem, which in a wider context it is also valid for any cycle problem. Another contribution has been the developed local search to handle large problems. The Branch-and-Cut algorithm includes new contributions in the separation algorithms of inequalities stemming from the cycle problem, in the separation loop, in the variables pricing, and in the calculation of the lower and upper bounds of the problem. At the same time, we have generalized for cycle problems the support graph shrinking techniques and procedures to speed up the exact separation algorithms for subcycle elimination constraints. The experiments carried out in large-sized instances, up to 7393 nodes, show that both algorithms achieve outstanding results, both in terms of the quality of solutions and in terms of the execution time., BERC.2014-2017 SEV-2013-0323 PID2019-104933GB-I00 MTM2015-65317-P
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Efficient meta-heuristics for spacecraft trajectory optimization

  • Shirazi, A.
Meta-heuristics has a long tradition in computer science. During the past few years, different types of meta-heuristics, specially evolutionary algorithms got noticeable attention in dealing with real-world optimization problems. Recent advances in this field along with rapid development of high processing computers, make it possible to tackle various engineering optimization problems with relative ease, omitting the barrier of unknown global optimal solutions due to the complexity of the problems. Following this rapid advancements, scientific communities shifted their attention towards the development of novel algorithms and techniques to satisfy their need in optimization. Among different research areas, astrodynamics and space engineering witnessed many trends in evolutionary algorithms for various types of problems. By having a look at the amount of publications regarding the development of meta-heuristics in aerospace sciences, it can be seen that a high amount of efforts are dedicated to develop novel stochastic techniques and more specifically, innovative evolutionary algorithms on a variety of subjects. In the past decade, one of the challenging problems in space engineering, which is tackled mainly by novel evolutionary algorithms by the researchers in the aerospace community is spacecraft trajectory optimization. Spacecraft trajectory optimization problem can be simply described as the discovery of a space trajectory for satellites and space vehicles that satisfies some criteria. While a space vehicle travels in space to reach a destination, either around the Earth or any other celestial body, it is crucial to maintain or change its flight path precisely to reach the desired final destination. Such travels between space orbits, called orbital maneuvers, need to be accomplished, while minimizing some objectives such as fuel consumption or the transfer time. In the engineering point of view, spacecraft trajectory optimization can be described as a black-box optimization problem, which can be constrained or unconstrained, depending on the formulation of the problem. In order to clarify the main motivation of the research in this thesis, first, it is necessary to discuss the status of the current trends in the development of evolutionary algorithms and tackling spacecraft trajectory optimization problems. Over the past decade, numerous research are dedicated to these subjects, mainly from two groups of scientific communities. The first group is the space engineering community. Having an overall look into the publications confirms that the focus in the developed methods in this group is mainly regarding the mathematical modeling and numerical approaches in dealing with spacecraft trajectory optimization problems. The majority of the strategies interact with mixed concepts of semi-analytical methods, discretization, interpolation and approximation techniques. When it comes to optimization, usually traditional algorithms are utilized and less attention is paid to the algorithm development. In some cases, researchers tried to tune the algorithms and make them more efficient. However, their efforts are mainly based on try-and-error and repetitions rather than analyzing the landscape of the optimization problem. The second group is the computer science community. Unlike the first group, the majority of the efforts in the research from this group has been dedicated to algorithm development, rather than developing novel techniques and approaches in trajectory optimization such as interpolation and approximation techniques. Research in this group generally ends in very efficient and robust optimization algorithms with high performance. However, they failed to put their algorithms in challenge with complex real-world optimization problems, with novel ideas as their model and approach. Instead, usually the standard optimization benchmark problems are selected to verify the algorithm performance. In particular, when it comes to solve a spacecraft trajectory optimization problem, this group mainly treats the problem as a black-box with not much concentration on the mathematical model or the approximation techniques. Taking into account the two aforementioned research perspectives, it can be seen that there is a missing link between these two schemes in dealing with spacecraft trajectory optimization problems. On one hand, we can see noticeable advances in mathematical models and approximation techniques on this subject, but with no efforts on the optimization algorithms. On the other hand, we have newly developed evolutionary algorithms for black-box optimization problems, which do not take advantage of novel approaches to increase the efficiency of the optimization process. In other words, there seems to be a missing connection between the characteristics of the problem in spacecraft trajectory optimization, which controls the shape of the solution domain, and the algorithm components, which controls the efficiency of the optimization process. This missing connection motivated us in developing efficient meta-heuristics for solving spacecraft trajectory optimization problems. By having the knowledge about the type of space mission, the features of the orbital maneuver, the mathematical modeling of the system dynamics, and the features of the employed approximation techniques, it is possible to adapt the performance of the algorithms. Knowing these features of the spacecraft trajectory optimization problem, the shape of the solution domain can be realized. In other words, it is possible to see how sensitive the problem is relative to each of its feature. This information can be used to develop efficient optimization algorithms with adaptive mechanisms, which take advantage of the features of the problem to conduct the optimization process toward better solutions. Such flexible adaptiveness, makes the algorithm robust to any changes of the space mission features. Therefore, within the perspective of space system design, the developed algorithms will be useful tools for obtaining optimal or near-optimal transfer trajectories within the conceptual and preliminary design of a spacecraft for a space mission. Having this motivation, the main goal in this research was the development of efficient meta-heuristics for spacecraft trajectory optimization. Regarding the type of the problem, we focused on space rendezvous problems, which covers the majority of orbital maneuvers, including long-range and short-range space rendezvous. Also, regarding the meta-heuristics, we concentrated mainly on evolutionary algorithms based on probabilistic modeling and hybridization. Following the research, two algorithms have been developed. First, a hybrid self adaptive evolutionary algorithm has been developed for multi-impulse long-range space rendezvous problems. The algorithm is a hybrid method, combined with auto-tuning techniques and an individual refinement procedure based on probabilistic distribution. Then, for the short-range space rendezvous trajectory optimization problems, an estimation of distribution algorithm with feasibility conserving mechanisms for constrained continuous optimization is developed. The proposed mechanisms implement seeding, learning and mapping methods within the optimization process. They include mixtures of probabilistic models, outlier detection algorithms and some heuristic techniques within the mapping process. Parallel to the development of algorithms, a simulation software is also developed as a complementary application. This tool is designed for visualization of the obtained results from the experiments in this research. It has been used mainly to obtain high-quality illustrations while simulating the trajectory of the spacecraft within the orbital maneuvers., La Caixa TIN2016-78365R PID2019-1064536A-I00 Basque Government consolidated groups 2019-2021 IT1244-19
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Contributions to Time Series Classification: Meta-Learning and Explainability

  • Abanda, A.
This thesis includes 3 contributions of different types to the area of supervised time series classification, a growing field of research due to the amount of time series collected daily in a wide variety of domains. In this context, the number of methods available for classifying time series is increasing, and the classifiers are becoming more and more competitive and varied. Thus, the first contribution of the thesis consists of proposing a taxonomy of distance-based time series classifiers, where an exhaustive review of the existing methods and their computational costs is made. Moreover, from the point of view of a non-expert user (even from that of an expert), choosing a suitable classifier for a given problem is a difficult task. The second contribution, therefore, deals with the recommendation of time series classifiers, for which we will use a meta-learning approach. Finally, the third contribution consists of proposing a method to explain the prediction of time series classifiers, in which we calculate the relevance of each region of a series in the prediction. This method of explanation is based on perturbations, for which we will consider specific and realistic transformations for the time series., BES-2016-076890
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Time Dynamics in Quantum Field Theory Systems

  • LIll, S.
In this doctoral thesis, we develop and investigate new mathematical tools that are intended to allow for a rigorous description of non–perturbative quantum field theory (QFT) dynamics. Here, the term QFT is to be understood as describing a quantum system with particle creation and annihilation that can, but does not need to, comply with special relativity. The tools aim at cases where a formal Hamiltonian exists but is ill–defined., Wilhelm Schuler-Stiftung Tübingen, DAAD (Deutscher Akademischer Austauschdienst)
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Various results concerning homogenisation of nematic liquid crystals

  • Ceuca, R.D.
In this thesis we present various results concerning homogenisation of nematic liquid crystals: two of them are in perforated domains, while the other one concerns rates of convergence for boundary homogenisation. The first work described in this thesis is a $\Gamma$-convergence result for the Landau-de Gennes model in $3D$ domains with connected perforations. The goal of the analysis is to find new terms in the energy functional that are independent of the gradient. The second result is an error estimate for a $2$D toy model used to describe rugosity effects in nematic liquid crystals via homogenisation problems, using once again the Landau-de Gennes model. The last problem is a local $L^2$-convergence result for a homogenisation problem in $\mathbb{R}^2$ with isolated perforations. Here we use the Oseen-Frank model, with the goal of finding new gradient-dependent terms in the energy functional. We start, in Chapter 1, with a brief introduction to nematic liquid crystals. We introduce two major variational models used to describe nematic liquid crystals: Landau-de Gennes (LdG) and Oseen-Frank (OF). For LdG theory, which uses $Q$-tensors as the order parameter, we present typical choices for each type of energy contribution (bulk, elastic and surface). For OF theory based on the order parameter $\textbf{n}\in\mathbb{S}^2$ we discuss the elastic energy, that depends on the director and its gradient. We then present a short summary of the main mathematical results obtained for LdG and OF. In Chapter 2, we analyse a homogenisation problem in $\mathbb{R}^3$ using the Landau-de Gennes model in which the perforations form a cubic microlattice. We assume a dillute regime, that is the volume of the cubic microlattice tends to $0$ as its characteristic length scale tends to $0$. The goal of this problem is to show that, given this geometrical setting, by choosing various types of surface energies one can obtain a new material in the limit of vanishing characteristic size of the microlattice. This material also behaves like a nematic liquid crystal, but now with different bulk coefficients. At the end of this chapter, we discuss a rate of convergence of the approximating surface energies to a homogenised term. In Chapter 3, we concentrate on achieving and improving error estimates in homogenisation problems, since they can give us crucial information for manufacturing processes. Here, we consider a simplified $2$D model in which we highlight how one could replace a rugose boundary with the imposed homeotropic alignment by a flat boundary with an effective alignment depending on the initial geometry of the rugosity. We are able to improve an $L^2$ error estimate for a class of linear nonhomogeneous Robin problems. In Chapter 4, we consider a general elastic energy in a $2$D case for the Oseen-Frank model in domains with isolated perforations. The goal of this study is to analyse how one could obtain a nematic liquid crystal with novel elastic properties via homogenisation procedures. Under suitable conditions we can analyse the $\mathbb{S}^1$-valued homogenisation problem via a scalar problem obtained through the lifting procedure. We also prove a local $L^2$ convergence result.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Supervised Learning in Time-dependent Environments with Performance Guarantees

  • Álvarez, V.
In practical scenarios, it is common to learn from a sequence of related problems (tasks). Such tasks are usually time-dependent in the sense that consecutive tasks are often significantly more similar. Time-dependency is common in multiple applications such as load forecasting, spam main filtering, and face emotion recognition. For instance, in the problem of load forecasting, the consumption patterns in consecutive time periods are significantly more similar since human habits and weather factors change gradually over time. Learning from a sequence tasks holds promise to enable accurate performance even with few samples per task by leveraging information from different tasks. However, harnessing the benefits of learning from a sequence of tasks is challenging since tasks are characterized by different underlying distributions. Most existing techniques are designed for situations where the tasks’ similarities do not depend on their order in the sequence. Existing techniques designed for timedependent tasks adapt to changes between consecutive tasks accounting for a scalar rate of change by using a carefully chosen parameter such as a learning rate or a weight factor. However, the tasks’ changes are commonly multidimensional, i.e., the timedependency often varies across different statistical characteristics describing the tasks. For instance, in the problem of load forecasting, the statistical characteristics related to weather factors often change differently from those related to generation. In this dissertation, we establish methodologies for supervised learning from a sequence of time-dependent tasks that effectively exploit information from all tasks, provide multidimensional adaptation to tasks’ changes, and provide computable tight performance guarantees. We develop methods for supervised learning settings where tasks arrive over time including techniques for supervised classification under concept drift (SCD) and techniques for continual learning (CL). In addition, we present techniques for load forecasting that can adapt to time changes in consumption patterns and assess intrinsic uncertainties in load demand. The numerical results show that the proposed methodologies can significantly improve the performance of existing methods using multiple benchmark datasets. This dissertation makes theoretical contributions leading to efficient algorithms for multiple machine learning scenarios that provide computable performance guarantees and superior performance than state-of-the-art techniques.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Spatio‑temporal modelling of high‑throughput phenotyping data

  • Pérez, D.M.
High throughput phenotyping (HTP) platforms and devices are increasingly used to characterise growth and developmental processes for large sets of plant genotypes. This dissertation is motivated by the need to accurately estimate genetic effects over time when analysing data from such HTP experiments. The HTP data we deal with here are characterised by phenotypic traits measured multiple times in the presence of spatial and temporal noise and a hierarchical organisation at three levels (populations, genotypes within populations, and plants within genotypes). The challenge is to balance efficient statistical models and com- putational solutions to deal with the complexity and dimensionality of the experimental data. To that aim, we propose two strategies. The first proposal divides the problem into two stages. The first stage (spatial model) focuses on correcting the phenotypic data for experimental design factors and spatial variation, while the second stage (hierarchical longitudinal model) aims to estimate the evolution over time of the genetic signal. The second proposal is to face the problem simultaneously (one-stage approach). That is, mod- elling the longitudinal evolution of the genetic effect on a given phenotypic trait while accounting for the temporal and spatial effects of environmental and design factors (spatio-temporal hierarchical model). We follow the same modelling philosophy throughout our work and propose multidimensional P-spline-based hierarchical approaches. We provide the user with appealing tools that take advantage of the sparse model matrices structure to reduce computational complexity. All our codes are publicly available on the R-package statgenHTP and https://gitlab.bcamath.org/dperez/htp_one_stage_approach. We illustrate the performance of our methods using spatio-temporal simulated data and data from the PhenoArch greenhouse platform at INRAE Montpellier and the outdoor Field Phenotyping platform at ETH Zürich. In the plant breeding context, we show how to extract new time-independent phenotypes for genomic selection purposes., MTM2017-82379-R BERC 2018-2021 BERC 2022-2025 SEV-2017-0718 CEX2021-001142-S/MICIN/AEI/10.13039/501100011033
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Easy-to-implement hp-adaptivity for non-elliptic goal-oriented problems

  • Caro, F.V.
The FEM has become a foundational numerical technique in computational mechanics and civil engineering since its inception by Courant in 1943 Courant1943. Originating from the Ritz method and variational calculus, the FEM was primarily employed to derive solutions for vibrational systems. A distinctive strength of the FEM is its capability to represent mathematical models through the weak variational formulation of PDE, facilitating computational feasibility even in intricate geometries. However, the search for accuracy often imposes a significant computational task. In the FEM, adaptive methods have emerged to balance the accuracy of solutions with computational costs. The $h$-adaptive FEM designs more efficient meshes by reducing the mesh size $h$ locally while keeping the polynomial order of approximation $p$ fixed (usually $p=1,2$). An alternative approach to the $h$-adaptive FEM is the $p$-adaptive FEM, which locally enriches the polynomial space $p$ while keeping the mesh size $h$ constant. By dynamically adapting $h$ and $p$, the $hp$-adaptive FEM achieves exponential convergence rates. Adaptivity is crucial for obtaining accurate solutions. However, the traditional focus on global norms, such as $L^2$ or $H^1$, might only sometimes serve the requirements of specific applications. In engineering, controlling errors in specific domains related to a particular QoI is often more critical than focusing on the overall solution. That motivated the development of GOA strategies. In this dissertation, we develop automatic GO $hp$-adaptive algorithms tailored for non-elliptic problems. These algorithms shine in terms of robustness and simplicity in their implementation, attributes that make them especially suitable for industrial applications. A key advantage of our methodologies is that they do not require computing reference solutions on globally refined grids. Nevertheless, our approach is limited to anisotropic $p$ and isotropic $h$ refinements. We conduct multiple tests to validate our algorithms. We probe the convergence behavior of our GO $h$- and $p$-adaptive algorithms using Helmholtz and convection-diffusion equations in one-dimensional scenarios. We test our GO $hp$-adaptive algorithms on Poisson, Helmholtz, and convection-diffusion equations in two dimensions. We use a Helmholtz-like scenario for three-dimensional cases to highlight the adaptability of our GO algorithms. We also create efficient ways to build large databases ideal for training DNN using $hp$ MAGO FEM. As a result, we efficiently generate large databases, possibly containing hundreds of thousands of synthetic datasets or measurements.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Damage identification in bridges combining deep learning and computational mechanic

  • Fernandez-Navamuel, A.
Civil infrastructures, such as bridges, are critical assets for society and the economy. Many of them have already reached their expected life and withstand loadings that exceed the design specifications. Besides, bridges suffer from various degradation mechanisms, including aging, corrosion, earthquakes, and, nowadays, the undeniable effect of climate change. This context has motivated an increasing interest in early detecting damage to prevent costly actions and dangerous failures. Structural Health Monitoring (SHM) consists of implementing effective strategies to continuously assess the health condition of structures using monitoring data collected by sensors. This dissertation focuses on the SHM problem of damage detection and identification. It is an ill-posed inverse problem that aims at inferring the health state of a structure from measurements of its response. The measurements include large amounts of noisy data affected by environmental and operational conditions, acquired with sensors of different nature. Solving such a multidisciplinary problem encompasses the use of applied mathematics, computational mechanics, and data science. In this dissertation, we exploit the potential of Deep Neural Networks in approximating complex inverse problems and employ computational parametrizations and the Finite Element Method to enrich the training phase by including damage scenarios. We explore two different approaches to the problem. In the first approach, we develop an outlier detection strategy to detect departures from the baseline condition. We only employ long-term monitoring data measured at the bridge during normal (healthy) operation. Starting from Principal Component Analysis (PCA) as a statistical data reconstruction technique, we design a specific Deep Autoencoder network that enhances PCA by adding residual connections to include nonlinear transformations. This architecture gains partial explainability by evaluating the contribution of nonlinearties over affine transformations in the reconstruction process. We also investigate the method performance when using local or global variables and evaluate the potential of combining both data sources in the damage detection task. In the second approach, we reach a higher level of damage identification by estimating its severity and location. The goal is to provide a suitable methodology for real full-scale applications that requires reasonable computational resources. We employ a calibrated computational parametrization to solve multiple Finite Element simulations under different damage scenarios. These synthetic scenarios enrich the training dataset of a Deep Neural Network that maps the response of the bridge with its health condition in terms of damage location and severity. Finally, we incorporate the effect of environmental and operational variability in the parametrization by applying a clustering algorithm to find representative samples among the entire dataset. We assume these samples cover most of the variability present in the data and consider them as starting points to generate synthetic training data. We apply the proposed methods to three main case study bridges with available monitoring data: the Beltran bridge in Mexico, and the Infante Dom Henrique bridge in Porto, and the Z24 bridge in Switzerland. Both structures resulted critical to validate and test the ability of the proposed methods and to demonstrate their applicability in the full-scale., This disseration has been possible thanks to the support received from: the European Union’s Horizon 2020 research and innovation program under the grant agreement No 769373 (FORESEE project) and the Marie Sklodowska-Curie grant agreement No 777778 (MATHROCKS); the Base Funding - UIDB/04708/2020 of the CONSTRUCT - Instituto de I&D em Estruturas e Constru¸c˜oes - funded by national funds through the FCT/MCTES (PIDDAC); the European Regional Development Fund (ERDF) through the Interreg V-A Spain-France-Andorra program POCTEFA 2014-2020 Project PIXIL (EFA362/19); the Spanish Ministry of Science and Innovation with references PID2019-108111RB-I00 (FEDER/AEI) and the “BCAM Severo Ochoa” accreditation of excellence (SEV-2017-0718); and the Basque Government through the BERC 2018-2021 program, the four Elkartek projects 3KIA (KK-2020/00049), EXPERTIA (KK-2021/00048), MATHEO (KK-2019-00085), and SIGZE (KK-2021/00095); the grant “Artificial Intelligence in BCAM number EXP. 2019/00432”, and the Consolidated Research Group MATHMODE (IT1294-19) given by the Department of Education.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Statistical Modelling for Recurrent Events in Sports Injury Research with Applications to Football InjuryData.

  • Zumeta, L.
Sports injuries stand as undesirable side effects of athletic participation, carrying serious consequencesfor athletes' health, their professional careers, and overall team performance. With the growing availability of data, there has been an increasing reliance on statistical models to monitor athletes' healthand mitigate injury risks.In this dissertation, our focus is on the statistical analysis of sports injury data, with an emphasis on the time-varying and recurrent nature of injury occurrences. We develop and assess suitable statistical modelling approaches to address specific research questions that arise in sports injury prevention research. We pursue three primary objectives: (a) identifying biomechanical risk factors using variableselection methods and shared frailty Cox models, (b) developing a flexible recurrent time-to-event approach to model the effects of training load on subsequent injuries, and (c) creating dedicated statistical tools through the open-source R software. These objectives are driven by interdisciplinary research, conducted in close collaboration with the Medical Services of Athletic Club, and are motivated by real-world applications. Namely, the work is based on three distinct data sets: the functional screening tests data, the external training load data, and the web-scraped football injury data. The statistical advancements developed contribute to ongoing efforts in sports injury prevention, providinginsights, methodologies, and accessible software implementations for sports medicine practitioners., This research was supported by the Spanish Ministry of Science and Innovation (MICINN) through the Severo Ochoa SEV-2017-0718 PRE2018- 084007 funding and the BCAM Severo Ochoa accreditation CEX2021-001142- S/MICIN/AEI/10.13039/501100011033; by the Basque Government through the BERC 2018-2021 and BERC 2022-2025 programs, and the PRE_2021_2_0029 funding; and by the AEI/FEDER, UE through the “S3M1P4R” PID2020-115882RB-I00 project.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Quantifying the impact of COVID-19 non-pharmaceutical interventions and vaccination using mathematical modeling and cost-effectiveness analysis

  • Estadilla, C.D.
In 2020, COVID-19 emerged as a global health crisis, prompting the widespread adoption of non-pharmaceutical interventions (NPIs) such as lockdowns, social distancing, and mask mandates. These measures, along with large-scale vaccination efforts, aimed to control virus transmission and reduce severe health outcomes among the elderly and those with comorbidities, albeit with considerable economic impacts. This dissertation examines the effectiveness and cost-effectiveness of public health measures against COVID-19, with a focus on the Basque Country—a region with a highly aging population. Through mathematical modeling and health economics, we assess the role of NPIs in reducing transmission, the impact of the disease in life expectancy, as well as the cost-effectiveness of vaccination, particularly given the spread of more transmissible variants. This work presents estimates of hospitalizations and deaths prevented by vaccination, using models fitted to two years of daily data, and compares these with findings from other settings. These insights inform a cost-effectiveness analysis from a healthcare payer perspective.
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Displacement processes with random shifting and their applications

  • Dahlenburg, M.
The thesis starts with a generalized model of Stochastic resetting that extends the classical concept through a random resetting amplitude. In particular there are two scenarios: independent and dependent Random amplitude stochastic resetting (RASR). A general analytical formalism of RASR is provided explicitly based on the first renewal picture, whereby the reduction to special cases is provided. The quantitative analysis is specified through the discussion of the mean value and the variance of the height in presence of a ballistic free propagation for Poissonian and constant pace resetting. These choices of resetting interval lengths are based on the geophysical application of sediment dynamics, in which the free propagation mimics the gradual increase (deposition) or decrease (erosion) of a sedimentation profile, whereas the resets represent an additional erosion mechanism. The latter mechanism could be seasonal (constant pace resetting) or random in time weather events such as extreme floods. The qualitative difference between independent and dependent RASR is that the latter class becomes stationary, whereas the former remains nonstationary. Apart from the above geophysical scenarios, this extension of the resetting dynamics can be applied to population dynamics interrupted by epidemics or crises-interrupted financial markets. All these applications correspond to the intermittent picture of a parent process with superimposed resetting statistic, which motivates the investigations of the RASR-model as a random search strategy. The following chapter investigates the search properties of the continuous time random walk (CTRW), which can be derived from the independent RASR-class. The problem leads to the derivation of a nonhomogeneous Wiener-Hopf equation that exactly calculates the mean first-passage time (MFPT). Apart from the fact, that the derived formalism may also be used for the indirect estimation of the jump-length PDF by assuming a given MFPT, it is shown by use of a specific jump-size PDF that the MFPT is infinite for symmetric random walks, but it becomes finite only for asymmetric jumps with a negative mean. Asymmetric jumps can be physically explained through an active movement as, e.g., chemotaxis or animal movement. However, for the quantitative description of molecular dynamics, symmetric CTRWs are presumed, for which the MFPT is not a proper quantity. In the subsequent chapters the author extended and modified these calculations for removing this failure of the MFPT for the investigation of symmetric random walks. In the fourth chapter of the thesis, a formalism for the calculation of the survival probability in first-passage time problems for symmetric random walks in semi-infinite space is provided. Therefore, a Sturm-Liouville system of equations may provide the unique solution, which is shown for a concrete example of jump-size densities in the discrete time scenario. In addition, the Sturm-Liouville system is also derived for the survival probability of a CTRW. The previous example of jump-size PDF is applied also in the continuous-time setting for arbitrary waiting-time distribution, that holds for Markovian and non-Markovian CTRW. In the last chapter of this thesis, the MFPT of a CTRW in the presence of a uniformly moving target is discussed. Formally, the MFPT depends on the whole jump-sizes and waiting-times distributions within this configuration. Thus, the simple picture that emerged for a fixed target, when the MFPT depends on the mean waiting time, only, is (almost always) broken. Nevertheless, for a uniformly departing target, the MFPT of a CTRW may fulfill the property to be independent of the entire waiting-time PDF for a certain class of jump-size densities, that is derived within the thesis. Moreover, it is shown that the average of the entire process, i.e., jumps plus uniform movement must be negative to provide a finite MFPT, which is not satisfied by symmetric jump-size PDFs in the presence of a departing target. However, for an arriving target, the MFPT is finite even for symmetric CTRWs. This has been observed for a paradigmatic setting for waiting-time PDF and jump-size density, for which the exact result is provided analytically., BERC 2018–2021, Predoc Severo Ochoa. PRE2018-084427
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Novel P-Spline-based approaches for curve and derivative estimation in longitudinal growth studies

  • Hernández, M.A.
Longitudinal data analysis plays a key role in understanding individual development over time. It helps identify important characteristics such as growth patterns, biological variations, and key milestones like growth spurts and the age at which they occur. Although several statistical models and software implementations are available to estimate growth curves and their derivatives, challenges remain due to the complex and dynamic nature of these processes. This dissertation explores a range of statistical methods for modeling growth based on repeated measurements from individuals. In particular, it highlights the use of penalized spline (P-spline) models, which have gained popularity for their flexibility in modeling smooth curves. Additionally, their connection with mixed-effects models allows us to take advantage of the methodological and computational tools in this setting for fitting P-spline models. We focus on three main classes of models for fitting individual growth trajectories and estimating their derivatives: (i) the double-penalty P-spline model, (ii) a family of semiparametric mixed-effects models, and (iii) modified versions of the SuperImposition by Translation and Rotation (SITAR) model. Each approach is discussed in terms of its theoretical background, practical implementation, and suitability for growth analysis. The proposed methods are evaluated through simulation studies and compared with alternative approaches to assess their performance. Our results indicate that the proposed models provide a slight improvement in curve fitting and derivative estimation. Additionally, we demonstrate the application of these methods using real longitudinal data from 125 young professional football players to analyze their growth and maturation processes over time., PID2020-115882RB-I00
BIRD. BCAM's Institutional Repository Data
Publicación Tesis doctoral (DoctoralThesis)

Pedestrian Dynamics Simulation Framework for Real-World Applications in Complex Public Environments

  • García, A.
This dissertation develops a robust pedestrian-dynamics simulation framework for real-life applications in complex public environments. Managing pedestrian flows has become increas- ingly critical in large cities and during mass events, where overcrowding can result in severe safety risks. The proposed framework enables the identification of high-density zones and the reproduction of complex pedestrian behaviors relevant for crowd management. In addition, the work addresses a contemporary challenge intensified by the COVID-19 pandemic: disease trans- mission in crowded indoor spaces. To this end, a novel infection-dynamics model is introduced, incorporating key aerosol-based transmission mechanisms such as anisotropic emission and in- halation, spatial diffusion, and viral decay in the air. The integrated pedestrian–infection model is applied to assess nosocomial infections in a hospital, incorporating real-life information on healthcare personnel work shifts. For pedestrian-dynamics modeling, this dissertation adopts the widely used Social Force Model. A comprehensive examination of the model is conducted to evaluate the impact of force formulations on pedestrian behavior, and to overcome its reduced capabilities by enabling the reproduction of key macroscopic empirical datasets while preserving collision-avoidance perfor- mance and emergent collective patterns reported in earlier implementations. Subsequently, the navigation in complex geometries is addressed by calculating time-independent vector fields to guide pedestrians towards intermediate and final targets within simulated environments. Fur- ther extensions are introduced to capture complex behaviors relevant in real-world environments, namely queue formation and route choice in multi-exit settings. By identifying queuing pedes- trians based on a cutoff radius and adjusting the dynamics of queuing agents, a wide range of formations are modeled, thus enhancing existing approaches. Building upon this queuing mecha- nism, a time-use efficiency algorithm is introduced to account for delays caused by queues and to replicate pedestrian decision-making. This method successfully reproduces route-choice behav- ior observed in independent experimental datasets featuring multiple exits, diverse geometries, and varying participant groups. The extended pedestrian-simulation methodology is ultimately applied to model spectator ingress at San Mamés Stadium. In this scenario, about 50,000 agents move around the stadium to reach one of 24 entry gates, generating diverse flow conditions, including unidirectional and bidirectional streams, sparse and dense regions, and queue formation. The model is used to evaluate the crowd-management protocols implemented during the 2024–2025 UEFA Europa League matches. The empirical access times at the turnstiles are accurately reproduced, yielding realistic pedestrian dynamics around the venue. This enables the identification of high-density areas associated with each management strategy and allows assessment of the impact of different arrival patterns and flow rates on local congestion levels. Finally, infection dynamics in crowds are investigated using a novel approach based on discretizing the diffusion equation via the smoothed-particle hydrodynamics (SPH) method. The anisotropic nature of human exhalation and inhalation is modeled by accounting for the relative face orientation between infected and susceptible individuals during direct transmission, which occurs within a 2 m radius of an infectious agent. Long-range indirect transmission is incorporated through spatial diffusion of viral particles in the air, together with exponential viral decay. After analyzing the influence of pedestrian mobility on the propagation dynamics, the integrated pedestrian–infection framework is applied to the 9B floor of Galdakao-Usansolo Hospital to study nosocomial COVID-19 transmission under routine daily conditions. The model assesses the impact of management interventions, such as reinforcing staff or controlling visits, on virus spread, assuming multiple sources of infection: patients, healthcare personnel, and visitors., BERC.2018-2021 Provincial Council of Bizkaia, 2023 Technology Transfer Program (M3OVE).