Dissertations / Theses on the topic 'Problemi di max'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 30 dissertations / theses for your research on the topic 'Problemi di max.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
DI, GIOVACCHINO STEFANO. "Approssimazione numerica structure-preserving di problemi stocastici di evoluzione." Doctoral thesis, Università degli Studi dell'Aquila, 2022. http://hdl.handle.net/11697/192064.
Full textPARENTE, ANGELO. "Ricerca di strutture speciali in problemi di programmazione lineare intera." Doctoral thesis, Università Politecnica delle Marche, 2013. http://hdl.handle.net/11566/242727.
Full textIn general, Integer Linear Programming problems are computationally hard to solve. The design of efficient algorithms for them often takes advantage from the analysis of the problem underlying mathematical structure. Starting from the problem of finding the maximum embedded network submatrix of a matrix with entries in {0,−1, 1}, this work deals with the equivalent problem of finding the maximum balanced induced subgraph of a signed graph (mbs, Max Balanced Subgraph). In particular, a new heuristic for the mbs problem, the Cycle-Contraction Heuristic (CCH), has been proposed. The algoritm is based on a graph trans- formation rule that progressively reduces the lengths of cycles, preserving at the same time the feasibility of solutions for the mbs problem. CCH turns out to be more effective of the state-of-the-art heuristics. The efficiency and the effectiveness of CCH can be further improved by means of new rules of data reduction, i.e, by a procedure that simplifies instances and/or decrease their size while preserving the optimal solution of the problem. In the second part of the work, a new exact approach for the mbs prob- lem has been proposed. Such method is based on a polynomial complexity transformation rule that turns a signed graphs into a simple 2-layer graph. The transformation estabilishes a strong connection between mbs and the well- known maximun independent set problem (mis) and allows to resort to a broad spectrum of (exact or heuristic) solution methods proposed in the literature for mis. Finally, the generalized counterpart on signed graphs of some well-known combinatorial problems have been investigated. In particular it has been proven that the k-coloring problem on a signed graph - the generalization on signed graph of the balancing problem and the generalization on a simple graph of the bipartization problem - is equivalent to mis problem on a k-layer graph, i.e., a simple graph obtained by generalizing the 2-layer transformation.
Pecoraro, Massimo. "Old and new problems in continuum structural materials." Doctoral thesis, Universita degli studi di Salerno, 2014. http://hdl.handle.net/10556/1469.
Full textThe PhD thesis in Mathematics titled "Old and New Problems in continuum structural masterials " is divided into two parts. In Part I ( Chapter I and II relative to the " Old Problems" ) is studied in particular the classic behavior of materials from the mechanical point of view. In Chapter I were determined tensors of stress and strain in a solid having the shape of a hollow cylindrical , that is not simply connected , when it does act on it a displacement field able to induce all six elementary distortions of Volterra in the case where the material constituting the solid is homogeneous , linearly elastic and transversely isotropic . Unlike Volterra considers an isotropic body, characterized by two elastic constants : (E and G) , presents in the constitutive relations of the material become five (A, C, F, L and N). The result obtained is that the functions of displacement [u1 (x,y,z), u2 (x,y,z), u3 (x,y,z) ] . meet the indefinite equations of elastic written using the five elastic constants , but as Volterra , they do not cancel the load over the entire border of the hollow cylinder . In other words, do not give rise to a real distortion in that the action of the shift functions do not carry the cylinder from a natural configuration to a spontaneous through an isothermal transformation in which the load boundary is null but only self- balanced... [edited by author]
XII n.s.
KOUHKOUH, HICHAM. "Alcuni problemi asintotici per equazioni di Hamilton-Jacobi-Bellman e applicazioni all'ottimizzazione globale." Doctoral thesis, Università degli studi di Padova, 2022. http://hdl.handle.net/11577/3444759.
Full textThe thesis deals with control theory and related topics both in deterministic and stochastic framework, with an emphasis on the analytical aspects and Hamilton-Jacobi equations. It is divided into four chapters. The first chapter deals with periodic homogenization and singular perturbations in deterministic control problems. The main results concern the convergence and characterization of the limit value function and the underlying optimal trajectories, using relaxed control limits. The second chapter is motivated by a recent algorithm in the context of Deep Learning, called "Deep relaxation of Stochastic Gradient Descent", and concerns singular perturbations for stochastic control problems where the new difficulty with respect to the existing literature lies in the unboundeness of the data. The asymptotic behaviors in this context were obtained after developing new probabilistic methods, together with an adaptation of the viscosity instruments to problems with unbounded data. Then the results were applied to the previously mentioned algorithm and to its extension which also involves the optimal control of the so-called learning-rate parameter. The third chapter is devoted to global optimization. It aims to construct a dynamic system that asymptotically reaches the global minimum of a given function. To do this, ideas from weak KAM theory and both deterministic and stochastic control problems are used. The main tools to prove convergence are occupational (random) measures and the asymptotic behavior of the solutions of Hamilton-Jacobi equations. The last chapter provides a new method with new results for the solvability of the ergodic equations of Hamilton-Jacobi-Bellman in the viscous case with unbounded and merely measurable ingredients. The latter appears in various asymptotic problems present in the literature and among those addressed in the previous chapters. The results also extend to ergodic Mean-Field Games which are studied in the same context.
De, Martino Giuseppe. "Multi-Value Numerical Modeling for Special Di erential Problems." Doctoral thesis, Universita degli studi di Salerno, 2015. http://hdl.handle.net/10556/1982.
Full textThe subject of this thesis is the analysis and development of new numerical methods for Ordinary Di erential Equations (ODEs). This studies are motivated by the fundamental role that ODEs play in applied mathematics and applied sciences in general. In particular, as is well known, ODEs are successfully used to describe phenomena evolving in time, but it is often very di cult or even impossible to nd a solution in closed form, since a general formula for the exact solution has never been found, apart from special cases. The most important cases in the applications are systems of ODEs, whose exact solution is even harder to nd; then the role played by numerical integrators for ODEs is fundamental to many applied scientists. It is probably impossible to count all the scienti c papers that made use of numerical integrators during the last century and this is enough to recognize the importance of them in the progress of modern science. Moreover, in modern research, models keep getting more complicated, in order to catch more and more peculiarities of the physical systems they describe, thus it is crucial to keep improving numerical integrator's e ciency and accuracy. The rst, simpler and most famous numerical integrator was introduced by Euler in 1768 and it is nowadays still used very often in many situations, especially in educational settings because of its immediacy, but also in the practical integration of simple and well-behaved systems of ODEs. Since that time, many mathematicians and applied scientists devoted their time to the research of new and more e cient methods (in terms of accuracy and computational cost). The development of numerical integrators followed both the scienti c interests and the technological progress of the ages during whom they were developed. In XIX century, when most of the calculations were executed by hand or at most with mechanical calculators, Adams and Bashfort introduced the rst linear multistep methods (1855) and the rst Runge- Kutta methods appeared (1895-1905) due to the early works of Carl Runge and Martin Kutta. Both multistep and Runge-Kutta methods generated an incredible amount of research and of great results, providing a great understanding of them and making them very reliable in the numerical integration of a large number of practical problems. It was only with the advent of the rst electronic computers that the computational cost started to be a less crucial problem and the research e orts started to move towards the development of problem-oriented methods. It is probably possible to say that the rst class of problems that needed an ad-hoc numerical treatment was that of sti problems. These problems require highly stable numerical integrators (see Section ??) or, in the worst cases, a reformulation of the problem itself. Crucial contributions to the theory of numerical integrators for ODEs were given in the XX century by J.C. Butcher, who developed a theory of order for Runge-Kutta methods based on rooted trees and introduced the family of General Linear Methods together with K. Burrage, that uni ed all the known families of methods for rst order ODEs under a single formulation. General Linear Methods are multistagemultivalue methods that combine the characteristics of Runge-Kutta and Linear Multistep integrators... [edited by Author]
XIII n.s.
DELLA, MARCA ROSSELLA. "Problemi di controllo in epidemiologia matematica e comportamentale." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2021. http://hdl.handle.net/11380/1237622.
Full textDespite major achievements in eliminating long-established infections (as in the very well known case of smallpox), recent decades have seen the continual emergence or re-emergence of infectious diseases (last but not least COVID-19). They are not only threats to global health, but direct and indirect costs generated by human and animal epidemics are responsible for significant economic losses worldwide. Mathematical models of infectious diseases spreading have played a significant role in infection control. On the one hand, they have given an important contribution to the biological and epidemiological understanding of disease outbreak patterns; on the other hand, they have helped to determine how and when to apply control measures in order to quickly and most effectively contain epidemics. Nonetheless, in order to shape local and global public health policies, it is essential to gain a better and more comprehensive understanding of effective actions to control diseases, by finding ways to employ new complexity layers. This was the main focus of the research I have carried out during my PhD; the products of this research are collected and connected in this thesis. However, because out of context, other problems I interested in have been excluded from this collection: they rely in the fields of autoimmune diseases and landscape ecology. We start with an Introduction chapter, which traces the history of epidemiological models, the rationales and the breathtaking incremental advances. We focus on two critical aspects: i) the qualitative and quantitative assessment of control strategies specific to the problem at hand (via e.g. optimal control or threshold policies); ii) the incorporation into the model of the human behavioral changes in response to disease dynamics. In this framework, our studies are inserted and contextualized. Hereafter, to each of them a specific chapter is devoted. The techniques used include the construction of appropriate models given by non-linear ordinary differential equations, their qualitative analysis (via e.g. stability and bifurcation theory), the parameterization and validation with available data. Numerical tests are performed with advanced simulation methods of dynamical systems. As far as optimal control problems are concerned, the formulation follows the classical approach by Pontryagin, while both direct and indirect optimization methods are adopted for the numerical resolution. In Chapter 1, within a basic Susceptible-Infected-Removed model framework, we address the problem of minimizing simultaneously the epidemic size and the eradication time via optimal vaccination or isolation strategies. A two-patches metapopulation epidemic model, describing the dynamics of Susceptibles and Infected in wildlife diseases, is formulated and analyzed in Chapter 2. Here, two types of localized culling strategies are considered and compared: proactive and reactive. Chapter 3 concerns a model for vaccine-preventable childhood diseases transmission, where newborns vaccination follows an imitation game dynamics and is affected by awareness campaigns by the public health system. Vaccination is also incorporated in the model of Chapter 4. Here, it addresses susceptible individuals of any age and depends on the information and rumors about the disease. Further, the vaccine effectiveness is assumed to be partial and waning over time. The last Chapter 5 is devoted to the ongoing pandemic of COVID-19. We build an epidemic model with information-dependent contact and quarantine rates. The model is applied to the Italian case and explicitly incorporates the progressive lockdown restrictions.
BOCCAFOLI, Matteo. "Assistenza sanitaria a domicilio: problemi multi-obiettivo d’instradamento di veicoli con bilanciamento di carico e fidelizzazione paziente-infermiera." Doctoral thesis, Università degli studi di Ferrara, 2013. http://hdl.handle.net/11392/2388893.
Full textBERTOCCHI, CARLA. "I Metodi Interior Point Incontrano le Reti Neurali: un'Applicazione al Deblurring di Immagini." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2020. http://hdl.handle.net/11380/1200562.
Full textThe aim of this thesis is to propose novel Deep Learning model to approach the image deblurring problem. This is a well known Inverse Problem that is usually reformulated as a regularized optimization problem, in which the objective function to be minimized is a sum of a data discrepancy measure with a regularization term. Also additional constraints can be imposed to incorporate a priori knowledge on the desired solution. In our work we consider smooth data-fidelity and regularization terms and we include constraints in the objective function by means of a logarithmic barrier. A proximal interior point method (IPM) is adopted to address the minimization step, in which the proximity operator is restricted only to the barrier function. The key issue of our proposed approach is the following: the regularization parameter, the barrier parameter and the step size needed in the iterations of the IPM are chosen by means of a particular Deep Learning strategy. In particular, we say that the IPM algorithm is unfolded in a neural network structure, whose training process merges with the optimization process. We used benchmarks image datasets to train the resulting neural network architecture and test our approach. Comparisons with standard gradient projection methods, with recent machine learning algorithms and also with other unfolded methods have been performed and the tests showed good performances and competitiveness of our approach.
DI, CREDICO GIULIA. "Metodo Energetico agli Elementi di Contorno per problemi di Elastodinamica 2D nel dominio del tempo." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2022. http://hdl.handle.net/11380/1265215.
Full textThis thesis deals with the application of the Energetic Boundary Element Method (BEM) for the resolution of elastodynamic problems in bidimensional unbounded domains, outside an open regular obstacle or external to a region with a closed Lipschitz boundary. Starting from the fundamental Green’s tensor, the differential problem is rewritten in terms of different Boundary Integral Equations (BIEs), suitable to solve problems equipped by Dirichlet or Neumann datum at the boundary. These BIEs are then set in a space-time weak form, based on energy arguments, and numerically solved by means of Energetic BEM. All the considered weak BIEs, once discretized, give rise to linear systems with lower triangular Toeplitz matrix, whose entries are quadruple space-time integrals. A consistent part of the thesis discusses the quadrature formulas employed to compute numerically the integrals in space variables on the boundary with high accuracy, and taking into account the characteristic space singularities: O(log(r)) for the single layer integral operator, O(1/r) for the double layer integral operator and O(1/r^2) for the hypersingular integral operator. Moreover, an accurate study of the integration domain in local variables allows to overcome the issues of the integration of peculiar step functions that feature all the integral kernels. A theoretical analysis of the indirect weak form with single layer operator has been executed, in order to prove properties of coercivity and continuity of the associated energetic bilinear operator, and numerous numerical results are presented to confirm the correctness and the effectiveness of the energetic BEM, showing in particular long time stability of the BIE solutions. In alternative to the uniform decomposition of the obstacle, I have taken into account different types of discretization that turn out to be useful, for instance, to catch the asymptotic behaviour of the single layer BIE solution at the endpoints of an open obstacle or at the corners of a polygonal closed arc. In particular, the solution of this BIE for a Dirichlet problem behaves like O(r^-1/2) at the extremes of a crack and like O(r^-w) near a corner, with the exponent w related to the amplitude of the angle. Meshes geometrically or algebraically refined at these critical points improve the convergence towards the solution: therefore, an in dept analysis of the error decay in energy norm is shown with respect to the type of refinement (h-version, p-version and hp-version have been in particular considered). The numerical results verify the theoretical slope of the estimated error for the various discretization method. Similar remarks and numerical experiments are also presented for Neumann problems, solved by indirect weak form depending on the hypersingular operator. Lastly, I take into account the following issue: when standard Lagrangian basis functions are considered, the BEM matrices are made by time-dependent blocks that are generally fully populated. The overall memory cost of the energetic BEM is O(M^2N), M and N being the number of space and time degrees of freedom, respectively. This can prevent the application of BEM to large scale realistic problems. Thus, in this thesis, a fast technique, based on the Adaptive Cross Approximation (ACA), is provided in order to get a low rank approximation of the time blocks, reducing drastically the number of the original entries to be evaluated. This procedure leads to a drop in the computational time, spent for the assembly and the resolution of the linear system, and in the memory storage requirements, which are generally relevant. The effectiveness of this strategy is theoretically proved for the single layer weak formulation and several numerical results are presented and discussed.
CRISCI, SERENA. "Proprietà spettrali dei metodi del gradiente per problemi di ottimizzazione con vincoli speciali." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2020. http://hdl.handle.net/11380/1200563.
Full textThe role of the steplength selection strategies in gradient methods has been widely investigated in the last decades. Starting from the pioneering paper of Barzilai and Borwein (1988), many efficient steplength rules have been designed, which contributed to make gradient-based approaches an effective tool for addressing large-scale optimization problems arising in many real-world applications. Most of these steplength selection rules have been developed in the unconstrained optimization framework, with the aim of exploiting some second-order information for achieving a fast annihilation of the gradient of the objective function. These steplength rules have been successfully applied also within gradient projection (GP) methods for constrained optimization, though, in this case, a detailed analysis on how the constraints may affect their spectral properties, as well as their formulation, has not been yet carried out. However, the convergence criteria for the GP method do not require restrictive hypothesis on the steplength parameter, provided that it is bounded away from zero and belongs to a predefined interval: this flexibility in the choice of the steplength allows to develop updating strategies aimed at optimizing the numerical behaviour, possibly in an inexpensive way. Motivated by these considerations, we analyse how, for quadratic programs, the original Barzilai-Borwein (BB) schemes are influenced by the presence of the feasible set. To this aim, we analyse their behaviour with respect to the spectrum of the Hessian of the objective function starting from the simpler case of box-constraints, and then moving to inspect the case of a more general feasible region expressed by a Single Linear equality constraint together with lower and upper Bounds (SLB). We propose modified versions of the BB rules (and their extensions), obtaining improvements of the gradient projection methods. Driven by this study on the BB rules, we extend the spectral analysis to the steplength updating strategy proposed by Roger Fletcher (2012) within the so-called Limited Memory Steepest Descent (LMSD) method. In particular, we combine the idea of the limited memory steplength approach with the gradient projection method for quadratic programming problems subject to box-constraints, investigating the possibility of modifying the original updating strategy in order to take into account the lower and the upper bounds in a suitable manner. The practical effectiveness of the proposed strategies has been tested in several numerical experiments on random large scale box-constrained and SLB quadratic problems, on some well known non quadratic problems and on a set of test problems arising from real-life applications.
Gentile, Chiara. "Metodo del gradiente coniugato per problemi ai minimi quadrati non lineari." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2016. http://amslaurea.unibo.it/12187/.
Full textCheng, Ai Lian Elena. "Algoritmi per la stima del rumore nella ricostruzione di immagini." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2014. http://amslaurea.unibo.it/7894/.
Full textXu, Fei <1984>. "Problemi e strategie della traduzione italiano-cinese --analisi della traduzione cinese di "I cinesi non muoiono mai"." Master's Degree Thesis, Università Ca' Foscari Venezia, 2015. http://hdl.handle.net/10579/5774.
Full textTONON, REMIS. "Due problemi di teoria analitica dei numeri: somme armoniche con i primi e distribuzione delle cifre di quozienti fra interi." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2020. http://hdl.handle.net/11380/1200579.
Full textIn the first part of this thesis we extend the results of the paper “Small values of signed harmonic sums” by Bettin, Molteni & Sanna (2018). There, the authors consider harmonic truncated series, where the summands can have a positive or a negative sign; using these objects to approximate any real value, they study the function that measures the precision of this approximation. In particular, they prove some bounds for this function in some specific ranges. In this thesis, we prove that the same result holds not only for the sequence of all natural numbers, but also for any subsequence that satisfies a growth hypothesis. Besides, in the case of the sequence of numbers that are the product of k distinct primes, where k is a fixed natural number, we obtain a significant improvement on the bounds for the approximating function. In the second part of this thesis, we improve the result of the paper “Probability of digits by dividing random numbers: a ψ and ζ functions approach” by Gambini, Mingari Scarpello & Ritelli (2012). The authors study there the distribution of the nth digit after the decimal point (in different bases) of all possible ratios between the first N natural numbers: they prove that it is not uniform, but it follows a law analogous to Benford’s one. In this thesis, we improve the error term found by the authors; besides, we study some further and different aspects and some variations of the problem, such as the uniformity of the formula.
MATOS, MENDES NILSON FELIPE. "Sistema di supporto alle decisioni basato sulla Ricerca Operativa per problemi nella logistica farmaceutica." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2021. http://hdl.handle.net/11380/1244339.
Full textHealthcare services are strongly dependent on the availability of equipment and medicines, as shortages can lead to treatments interruptions, reduced capacity, or undesirable delays. In the last decades, centralized group purchasing organizations, coupled with an outsourced pharmaceutical logistic, have replaced traditional approaches to avoid shortages and associated negative effects. In Italy, this process started in the 1990s with a regionalization conducted by the Italian National Health Service. To make centralization strategy works, however, a good integration between warehouses and delivery infrastructure is fundamental. This means taking many decisions at all managerial levels. As these decisions are hard to be evaluated by hand, a computational tool becomes essential. In this thesis, we present algorithms and development processes used to create a decision support system for a pharmaceutical logistic company specialized in storage and distribution of pharmaceutical products in Italy. In the first chapter, the software conception and implementation are presented, including details about technology integration. In the second chapter, the transportation part of the system is presented, with a focus on the computational approach to solve two closely related problems, a rich vehicle routing problem and a truck and driver scheduling problem. In the third chapter, we present a storage allocation problem that has special constraints associated with the pharmaceutical logistic, and an Iterated Local Search (ILS) based algorithm to solve it. In conclusion, some possible system improvements and future research directions are shown. Additionally, the appendix contains two chapters that describe the results obtained in parallel researches developed by the author. The first appendix presents an Adaptative Large Neighbourhood Search heuristic combined with a Set Partitioning model to solve a multiobjective dial-a-flight problem. In this problem, a heterogeneous fleet of airplanes must be routed to carry passengers to a required destination. The objective is to minimize user inconvenience (measured by delays and intermediary stops) and costs. Each airplane has different speed, fuel consumption, capacity, and costs. The problem contains some hard-operational constraints such as airplane maximum weight (including passengers), safety, unavailability of fuel in some airports and time windows. The second appendix proposes four mathematical models and an ILS based heuristic to optimize a scheduling problem with position-dependent deterioration and maintenance activities. In this problem, a set of jobs must be scheduled on a set of parallel unrelated machines in order to minimize the makespan. Each job has an individual runtime and causes a deterioration on the machine that makes the runtimes of the next jobs rise by a cumulative factor. Maintenances, which have significant runtimes, can be scheduled between two jobs, making the machine recover its full performance. Overall, the contributions of this thesis lie in the proposal of algorithms and software to solve very complex routing and scheduling problems deriving from real-world decision problems.
SANNA, DANIELA. "Il pensiero algebrico:questioni di natura epistemologica e didattica." Doctoral thesis, Università degli Studi di Cagliari, 2016. http://hdl.handle.net/11584/266669.
Full textPIZZUTI, ANDREA. "Pricing-based primal and dual bounds for selected packing problems." Doctoral thesis, Università Politecnica delle Marche, 2020. http://hdl.handle.net/11566/273679.
Full textSeveral optimization problems ask for finding solutions which define packing of elements while maximizing (minimizing) the objective function. Solving some of these problems can be extremely challenging due to their innate complexity and the corresponding integer formulations can be not suitable to be solved on instances of relevant size. Thus, clever techniques must be devised to achieve good primal and dual bounds. A valid way is to rely on pricing-based algorithms, in which solution components are generated by calling and solving appropriate optimization subproblems. Two main exponents of this group are the delayed column generation (CG) procedure and the sequential value correction (SVC) heuristic: the former provides a dual bound based on the generation of implicit columns by examining the shadow prices of hidden variables; the latter explores the primal solution space by following the dynamic of approximate prices. In this thesis we focus on the application of SVC and CG techniques to find primal and dual bounds for selected packing problems. In particular, we study problems belonging to the family of cutting and packing, where the classical BIN PACKING and CUTTING STOCK are enriched with features derived from the real manufacturing environment. Moreover, the MAXIMUM γ-QUASI-CLIQUE problem is taken into account, in which we seek for the induced γ-quasi-clique with the maximum number of vertices. Computational results are given to assess the performance of the implemented algorithms.
Furieri, Luca. "Teoria del consenso e applicazione al problema del coordinamento del moto di robot." Bachelor's thesis, Alma Mater Studiorum - Università di Bologna, 2014. http://amslaurea.unibo.it/7556/.
Full textVEZZALI, DARIO. "Ottimizzazione integrata e sistemi a supporto delle decisioni per problemi di consegne e servizi a domicilio." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2023. https://hdl.handle.net/11380/1298349.
Full textThis research addresses the class of attended home delivery and service problems, which have been widely studied in the last two decades. In addition, The recent COVID-19 pandemic has boosted the interest in attended home delivery and service, with a significant increase in terms of global demand and a still lasting effect on people's habits. From an operational perspective, such problems require the customer to be present at home when the goods are delivered or the service is executed. Typically, the service provider and the customer agree on a particular time window for the delivery of goods or the execution of a service. The purpose of this research is to review the state of the art for attended home delivery and service problems, and study specific real-world applications in gas and water distribution, as well as in the context of global service providers. In particular, the solution of real-world optimization problems through integrated decision support systems, which rely on mathematical formulations plus additional modules, is investigated. This requires a careful problem definition, in order to clearly state the objective function and the main constraints of the application at hand, followed by the implementation in an exact or heuristic fashion, and ended with several computational experiments aimed at producing valuable solutions. This iterative process implies a preliminary real-data collection and preparation, which has to be performed as carefully, in order to compute all the relevant information that occurs in the decision-making process. The proposed methodology integrates classic techniques of operations research with machine learning, to predict missing information for future periods, multi-criteria decision analysis, to define and weight the multiple factors that determine a complex decision, and engineering economics, to evaluate a project from a financial perspective.
FRANCHINI, Giorgia. "Selezione degli iperparametri nei metodi di ottimizzazione stocastica." Doctoral thesis, Università degli studi di Modena e Reggio Emilia, 2021. http://hdl.handle.net/11380/1237616.
Full textIn the context of deep learning, the computational more expensive phase is the full training of the learning algorithm. Indeed the design of a new learning algorithm requires an extensive numerical investigation with the execution of a significant number of experimental trials. A fundamental aspect in designing a suitable learning algorithm is the selection of the hyperparameters (parameters that are not trained during the learning process), in a static or adaptive way. The aim of this thesis is to investigate the hyperparameters selection strategies on which standard machine learning algorithms are designed. In particular, we are interested in the different techniques to select the parameters related to the stochastic gradient methods used for training the machine learning methodologies. The main purposes that motivate this study are the improvement of the accuracy (or other metrics suitable for evaluating the inspected methodology) and the acceleration of the convergence rate of the iterative optimization schemes. To achieve these purposes, the analysis has mainly focused on the choice of the fundamental parameters (hyperparameters) in the stochastic gradient methods: the steplength, the minibatch size and the potential adoption of variance reduction techniques. In a first approach we considered separately the choice of steplength and minibatch size; then, we presented a technique that combines the two choices. About the steplength selection, we propose to tailor for the stochastic gradient iteration the steplength selection adopted in the full-gradient method known as Limited Memory Steepest Descent method. This strategy, based on the Ritz-like values of a suitable matrix, enables to give a local estimate of the inverse of the local Lipschitz parameter. Regarding the minibatch size the idea is to increasing dynamically, in an adaptive manner (based on suitable validation tests), this size. The experiments show that this training strategy is more efficient (in terms of time and costs) compared with the approaches available in literature. We combine the two parameter choices (steplength and minibatch size) in an adaptive scheme without introducing line search techniques, while the possible increase of the size of the subsample used to compute the stochastic gradient enables to control the variance of this direction. In the second part of the thesis, we introduce an Automatic Machine Learning (AutoML) technique to set these parameters. In particular, we propose a low-cost strategy to predict the accuracy of the learning algorithm, based only on its initial behavior. The initial and final accuracies observed during this beforehand process are stored in a database and used as training set of a Support Vector Machines learning algorithm. The aim is to predict the accuracy of a learning methodology, given its accuracy on the initial iterations of its learning process. In other word, by a probabilistic exploration of the hyperparameter space, we are able to find the setting providing optimal accuracies at a quite low cost. An extensive numerical experimentation was carried out involving convex and non-convex functions (in particular Convolutional Neural Networks). For the numerical experiments several datasets well known in literature have been used, for different problems such as: classification, segmentation, regression. Finally, a computational study is carried out to extend the proposed approaches to other methods, such as: Momentum, ADAM, SVRG. In conclusion, the contribution of the thesis consists in providing useful ideas about an effective and inexpensive selection of the hyperparameters in the class of the stochastic gradient methods.
Pinton, Stefano. "Regularity of the dbar-Neumann problem and the Green operator." Doctoral thesis, Università degli studi di Padova, 2012. http://hdl.handle.net/11577/3426289.
Full textQuesta tesi tratta la regolarità del problema dibar-Neumann e del sistema di Cauchy-Riemann tangenziale. Nel capitolo 1 si discute delle stime di compattezza. Si prova qui che esse sussistono in presenza della"(CR P-property)". Il nostro approccio si basa su una stima di base stabilita da T.V. Khanh che migliora risultati precedenti di A. Nicoara. Il Capitolo 2 tratta la regolarità del problema dibar-Neumann in assenza di stime di compattezza. Il primo approccio consiste nella condizione di ``buon campo vettore T" o ``buona funzione definitoria r. Da un lato questa condizione dà regolarità; dall'altro, essa è soddisfatta quando c'è una funzione definitoria plurisubarmonica r. La condizione di campo vettore è stata sostituita da una più debole condizione di tipo "moltiplicatore". Noi riprendiamo questa condizione e ne diamo una versione "quantificata". Il capitolo 3 tratta l'ipoellitticità con perdita di derivate sia per campi vettoriali sia per somme di quadrati. Il nostro contributo consiste nel trattare campi vettoriali modificati da campi di tipo esponenziale anzichè, classicamente, di tipo finito.
Violin, Alessia. "Mathematical programming approaches to pricing problems." Doctoral thesis, Università degli studi di Trieste, 2014. http://hdl.handle.net/10077/10863.
Full textThere are many real cases where a company needs to determine the price of its products so as to maximise its revenue or profit. To do so, the company must consider customers’ reactions to these prices, as they may refuse to buy a given product or service if its price is too high. This is commonly known in literature as a pricing problem. This class of problems, which is typically bilevel, was first studied in the 1990s and is NP-hard, although polynomial algorithms do exist for some particular cases. Many questions are still open on this subject. The aim of this thesis is to investigate mathematical properties of pricing problems, in order to find structural properties, formulations and solution methods that are as efficient as possible. In particular, we focus our attention on pricing problems over a network. In this framework, an authority owns a subset of arcs and imposes tolls on them, in an attempt to maximise his/her revenue, while users travel on the network, seeking for their minimum cost path. First, we provide a detailed review of the state of the art on bilevel pricing problems. Then, we consider a particular case where the authority is using an unit toll scheme on his/her subset of arcs, imposing either the same toll on all of them, or a toll proportional to a given parameter particular to each arc (for instance a per kilometre toll). We show that if tolls are all equal then the complexity of the problem is polynomial, whereas in case of proportional tolls it is pseudo-polynomial. We then address a robust approach taking into account uncertainty on parameters. We solve some polynomial cases of the pricing problem where uncertainty is considered using an interval representation. Finally, we focus on another particular case where toll arcs are connected such that they constitute a path, as occurs on highways. We develop a Dantzig-Wolfe reformulation and present a Branch-and-Cut-and-Price algorithm to solve it. Several improvements are proposed, both for the column generation algorithm used to solve the linear relaxation and for the branching part used to find integer solutions. Numerical results are also presented to highlight the efficiency of the proposed strategies. This problem is proved to be APX-hard and a theoretical comparison between our model and another one from the literature is carried out.
Un problème classique pour une compagnie est la tarification de ses produits à vendre sur le marché, de façon à maximiser les revenus. Dans ce contexte, il est important que la société prenne en compte le comportement de ses clients potentiels, puisque si le prix est trop élevé, ils peuvent décider de ne rien acheter. Ce problème est communément connu dans la littérature comme un problème de tarification ou "pricing". Une approche de programmation biniveau pour ce problème a été introduite dans les années 90, révélant sa difficulté. Cependant, certains cas particuliers peuvent être résolus par des algorithmes polynomiaux, et il y a encore de nombreuses questions ouvertes sur le sujet. Cette thèse de doctorat porte sur les propriétés mathématiques des problèmes de tarification, fixant l’objectif de déterminer différentes formulations et méthodes de résolution les plus efficaces possibles, en se concentrant sur les problèmes appliqués aux réseaux de différents types. Dans les problèmes de tarification sur réseau, nous avons deux entités : une autorité qui possède un certain sous-ensemble d’arcs, et impose des péages, avec l’intention de maximiser les revenus provenant de celle-ci, et des utilisateurs qui choisissent leur chemin de moindre coût sur l’ensemble du réseau. Dans la première partie de la thèse une analyse détaillée de l’état de l’art sur les problèmes de tarification biniveau est présentée, suivie, dans la deuxième partie, par une analyse de cas particuliers polynomiaux. En particulier, nous considérons le cas où l’autorité utilise un péage unitaire sur son sous-ensemble d’arcs, soit en choisissant le même péage sur chaque arc, soit en choisissant un péage proportionnel à un paramètre donné pour chaque arc (par exemple, un péage par kilomètre). Dans le premier cas de péages égaux, il est démontré que la complexité du problème est polynomiale, tandis que dans le second cas de péages proportionnels, elle est pseudo-polynomiale. Ensuite, nous présentons une première approche d’optimisation robuste pour les problèmes de tarification sur réseau, de manière à inclure de l’incertitude sur la valeur exacte des paramètres dans le modèle, qui est typique dans les problèmes réels. Cette incertitude est représentée en utilisant des intervalles pour les paramètres et nous proposons, pour certains cas, des algorithmes de résolution polynomiaux. La troisième et dernière partie de la thèse concerne un cas difficile, le problème de tarification sur réseau dans lequel les arcs sont connectés de manière à constituer un chemin, comme c’est le cas pour les autoroutes. Initialement, nous prouvons que ce problème est APX-dur, renforçant le résultat connu jusqu’à maintenant. Ensuite, nous présentons des nouvelles formulations plus fortes, et en particulier, nous développons une reformulation de type Danztig-Wolfe, résolue par un algorithme de Branch-and-Cut-and-Price. Enfin, nous proposons différentes stratégies pour améliorer les performances de l’algorithme, pour ce qui concerne l’algorithme de génération de colonnes utilisé pour résoudre la relaxation linéaire, et pour ce qui concerne la résolution du problème avec variables binaires. Les résultats numériques complètent les résultats théoriques, en mettant en évidence l’efficacité des stratégies proposées.
Un classico problema aziendale è la determinazione del prezzo dei prodotti da vendere sul mercato, in modo tale da massimizzare le entrate che ne deriveranno. In tale contesto è importante che l’azienda tenga in considerazione il comportamento dei propri potenziali clienti, in quanto questi ultimi potrebbero ritenere che il prezzo sia troppo alto e decidere dunque di non acquistare. Questo problema è comunemente noto in letteratura come problema di tariffazione o di “pricing”. Tale problema è stato studiato negli anni novanta mediante un approccio bilivello, rivelandone l’alta complessità computazionale. Tuttavia alcuni casi particolari possono essere risolti mediante algoritmi polinomiali, e ci sono sono ancora molte domande aperte sull’argomento. Questa tesi di dottorato si focalizza sulle proprietà matematiche dei problemi di tariffazione, ponendosi l’obiettivo di determinarne formulazioni e metodi risolutivi più efficienti possibili, concentrandosi sui problemi applicati a reti di vario tipo. Nei problemi di tariffazione su rete si hanno due entità: un’autorità che possiede un certo sottoinsieme di archi e vi impone dei pedaggi, con l’intento di massimizzare le entrate che ne derivano, e gli utenti che scelgono il proprio percorso a costo minimo sulla rete complessiva (a pedaggio e non). Nella prima parte della tesi viene affrontata una dettagliata analisi dello stato dell’arte sui problemi di tariffazione bilivello, seguita, nella seconda parte, dall’analisi di particolari casi polinomiali del problema. In particolare si considera il caso in cui l’autorità utilizza uno schema di pedaggio unitario sul suo sottoinsieme di archi, imponendo o lo stesso pedaggio su ogni arco, o un pedaggio proporzionale a un dato parametro relativo ad ogni arco (ad esempio un pedaggio al chilometro). Nel primo caso di pedaggi uguali, si dimostra che la complessità del problema è polinomiale, mentre nel secondo caso di pedaggi proporzionali è pseudo-polinomiale. In seguito viene affrontato un approccio di ottimizzazione robusta per alcuni problemi di tariffazione su rete, in modo da includere nei modelli un’incertezza sul valore esatto dei parametri,tipica dei problemi reali. Tale incertezza viene rappresentata vincolando i parametri in degli intervalli e si propongono, per alcuni casi, algoritmi risolutivi polinomiali. La terza e ultima parte della tesi riguarda un caso computazionalmente difficile, in cui gli archi tariffabili sono connessi in modo tale da costituire un cammino, come avviene per le autostrade. Inizialmente si dimostra che tale problema è APX-hard, rafforzando il risultato finora conosciuto. In seguito si considerano formulazioni piùforti, e in particolare si sviluppa una riformulazione di Danztig-Wolfe, risolta tramite un algoritmo di Branch-and-Cut-and-Price. Infine si propongono diverse strategie per migliorare le performance dell’algoritmo, sia per quanto riguarda l’algoritmo di generazione di colonne utilizzato per risolvere il rilassamento lineare, sia per quanto riguarda la risoluzione del problema con variabili binarie. Risultati numerici complementano quelli teorici ed evidenziano l’efficacia delle strategie proposte.
XXV Ciclo
1985
Brigandì, Camilla. "Utilizzo della omologia persistente nelle reti neurali." Bachelor's thesis, Alma Mater Studiorum - Università di Bologna, 2022.
Find full textCISTERNINO, MARCO. "A parallel second order Cartesian method for elliptic interface problems and its application to tumor growth model." Doctoral thesis, Politecnico di Torino, 2012. http://hdl.handle.net/11583/2497156.
Full textBONENTI, FRANCESCA. "La matematica come occasione e stimolo per la formulazione di un giudizio critico." Doctoral thesis, Università degli studi di Bergamo, 2013. http://hdl.handle.net/10446/28639.
Full textRICCA, Federica. "Metodi Combinatori per Problemi di Aggregazione e Partizione con Criteri Multipli." Doctoral thesis, 1998. http://hdl.handle.net/11573/618259.
Full textORLANDI, Giandomenico. "Alcuni problemi variazionali geometrici suggeriti dalla Fisica." Doctoral thesis, 1997. http://hdl.handle.net/11562/347430.
Full textWe deal with some variational problems with geometrical constraints of homological or homotopical type, or in fiber bundle ambients. These models arise in condensed matter Physics to describe phase transitions at low temperature (Allen-Cahn or Ginzburg-Landau models), and also in particle Physics (abelian Higgs models). We characterize the asymptotic behavior of equilibrium configurations, corresponding to global minimizers of the involved functionals, whose energy concentrate on codimension one or two minimal surfaces.
PEZZA, Laura. "Su un modello di Hele-Shaw dipendente dalla temperatura." Doctoral thesis, 1996. http://hdl.handle.net/11573/484674.
Full textWe constructed a model for the injection of a fluid in a Hele-Shaw cell and we study the properties of existence and uniqueness of the differential problem solution. These equations are of parabolic and transport kind, with an integral differential condition on the interface.
SBRANA, ALESSANDRO. "Faculty Development Centri di Professionalità Accademica (CPA)." Doctoral thesis, 2018. http://hdl.handle.net/11393/251175.
Full textGUIDI, Arianna. "Il reato a concorso necessario improprio." Doctoral thesis, 2018. http://hdl.handle.net/11393/251080.
Full text