| 1. |
- Abdalmoaty, Mohamed, 1986-
(författare)
-
Identification of Stochastic Nonlinear Dynamical Models Using Estimating Functions
- 2019
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Data-driven modeling of stochastic nonlinear systems is recognized as a very challenging problem, even when reduced to a parameter estimation problem. A main difficulty is the intractability of the likelihood function, which renders favored estimation methods, such as the maximum likelihood method, analytically intractable. During the last decade, several numerical methods have been developed to approximately solve the maximum likelihood problem. A class of algorithms that attracted considerable attention is based on sequential Monte Carlo algorithms (also known as particle filters/smoothers) and particle Markov chain Monte Carlo algorithms. These algorithms were able to obtain impressive results on several challenging benchmark problems; however, their application is so far limited to cases where fundamental limitations, such as the sample impoverishment and path degeneracy problems, can be avoided.This thesis introduces relatively simple alternative parameter estimation methods that may be used for fairly general stochastic nonlinear dynamical models. They are based on one-step-ahead predictors that are linear in the observed outputs and do not require the computations of the likelihood function. Therefore, the resulting estimators are relatively easy to compute and may be highly competitive in this regard: they are in fact defined by analytically tractable objective functions in several relevant cases. In cases where the predictors are analytically intractable due to the complexity of the model, it is possible to resort to {plain} Monte Carlo approximations. Under certain assumptions on the data and some conditions on the model, the convergence and consistency of the estimators can be established. Several numerical simulation examples and a recent real-data benchmark problem demonstrate a good performance of the proposed method, in several cases that are considered challenging, with a considerable reduction in computational time in comparison with state-of-the-art sequential Monte Carlo implementations of the ML estimator.Moreover, we provide some insight into the asymptotic properties of the proposed methods. We show that the accuracy of the estimators depends on the model parameterization and the shape of the unknown distribution of the outputs (via the third and fourth moments). In particular, it is shown that when the model is non-Gaussian, a prediction error method based on the Gaussian assumption is not necessarily more accurate than one based on an optimally weighted parameter-independent quadratic norm. Therefore, it is generally not obvious which method should be used. This result comes in contrast to a current belief in some of the literature on the subject. Furthermore, we introduce the estimating functions approach, which was mainly developed in the statistics literature, as a generalization of the maximum likelihood and prediction error methods. We show how it may be used to systematically define optimal estimators, within a predefined class, using only a partial specification of the probabilistic model. Unless the model is Gaussian, this leads to estimators that are asymptotically uniformly more accurate than linear prediction error methods when quadratic criteria are used. Convergence and consistency are established under standard regularity and identifiability assumptions akin to those of prediction error methods.Finally, we consider the problem of closed-loop identification when the system is stochastic and nonlinear. A couple of scenarios given by the assumptions on the disturbances, the measurement noise and the knowledge of the feedback mechanism are considered. They include a challenging case where the feedback mechanism is completely unknown to the user. Our methods can be regarded as generalizations of some classical closed-loop identification approaches for the linear time-invariant case. We provide an asymptotic analysis of the methods, and demonstrate their properties in a simulation example.
|
|
| 2. |
- Colarieti Tosti, Massimiliano
(författare)
-
Theory of Crystal Fields and Magnetism of f-electron Systems
- 2004
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- A parameter free approach for the calculation of the crystal field splitting of the lowest Russel-Saunders J-multiplet in f-electron systems has been developed and applied to selected compounds. The developed theory is applicable to general symmetries and is based on symmetry constrained density functional theory calculations in the local density or in the generalised gradient approximation.The magnetocrystalline anisotropy of Gd has been analysed. It has been shown that the peculiar orientation of the easy axis of magnetisation is consistent with an S-ground state. Further, the temperature dependence of the easy axis of magnetisation has been investigated and it has been shown that the temperature driven reduction of the effective magnetisation is the principal mechanism responsible for it.A new method has been developed that allows for theoretical studies of the electronic structure and total energy of elements and compounds in an intermediate valence regime. The method combines model and first-principles band structure calculations, therefore being accurate and computationally efficient. It has been applied to Yb metal under pressure obtaining a remarkable agreement with experimental observations for the equation of state and the x-ray absorption spectroscopy.
|
|
| 3. |
- Delcey, Mickaël G., 1988-
(författare)
-
Extending the Reach of Accurate Wavefunction Methods
- 2015
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Multiconfigurational quantum chemistry methods, and especially the multiconfigurational self-consistent field (MCSCF) and multireference perturbation theory (MRPT2), are powerful tools, particularly suited to the accurate modeling of photochemical processes and transition metal catalysis. However, they are limited by their high computational cost compared to other methods, especially density functional theory. Moreover, there are areas where they would be expected to perform well, but where they are not applied due to lack of experience.This thesis addresses those issues. First, the efficiency of the Cholesky decomposition approximation to reduce the cost of MCSCF and MRPT2 without sacrificing their accuracy is demonstrated. This then motivates the extension of the Cholesky approximation to the computation of MCSCF nuclear gradients, thus strongly improving the ability to perform MCSCF non-adiabatic molecular dynamics. Typically, a tenfold speed-up is observed allowing dynamic simulation of larger systems or over longer times.Finally, multiconfigurational methods are applied to the computation of X-ray spectra of transition metal complexes. The importance of the different parameters in the calculation is systematically investigated, laying the base for wider applications of those accurate methods in the modeling of X-ray spectroscopy. A tool to analyze the resulting spectrum in terms of molecular orbitals is also presented, strengthening the interplay between theory and experiments.With these developments and other significant ones that have happened in recent years, multiconfigurational methods can now reach new grounds and contribute to important new discoveries
|
|
| 4. |
|
|
| 5. |
- Dussauge, Isabelle, 1978-
(författare)
-
Technomedical Visions : Magnetic Resonance Imaging in 1980s Sweden
- 2008
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The medical imaging technology called MRI (magnetic resonance imaging) stems from a blind measurement technology which was further developed in research and practice to enable seeing into the inner body. Vision with MRI was open-ended, and it was developed and tamed in a context of fragmented medical perspectives on the body and on technology. "Technomedical Visions" addresses the formation of MRI’s specific visualities in the first decade of its introduction in Sweden. The purpose of this dissertation is to explore how vision with MRI has been constructed in practice in relation to existing ways of knowing the body within medicine. Dussauge investigates first the early decisions that led to a national evaluation of MRI technology in the mid-1980s in Sweden. Then she addresses the shaping of MRI’s quantitative visuality in the practices of radiology, psychiatry and the laboratory, with focus on microhistories at St. Göran’s Hospital, Karolinska Institutet, Uppsala University Hospital, and Lund University. Dussauge shows that whereas authorities’ early decisions momentarily defined MRI as a radiological tool for immediate clinical use and evaluation, a crucial part of MRI’s introduction was the work conducted by MRI-users. These researchers from a range of scientific and medical disciplines performed, over time, a multitude of shapings of MRI’s vision. This studies shows how MRI was made congruent with existing technomedical gazes. The novel MRI gaze was made intelligible within cross-referential networks, and researchers reproduced technomedicine’s existing gazes both in the production, optimization and interpretation of MRI representations. Technomedical time frames, epistemologies and definitions of the normal and the pathological were reproduced and sometimes, re-cast, in the shaping of MRI in practice. This study also demonstrates that anatomy recurrently worked as an underlying frame for the exploration and production of MRI visions. Anatomy’s material visuality provided a site for the production of novel facts at the intersection of existing gazes. Through the practices of shaping MRI gazes, anatomy was systematically remediated, reproduced and reconfigured.
|
|
| 6. |
- Geschwind, Lars, 1970-
(författare)
-
Stökiga studenter : Social kontroll och identifikation vid universiteten i Uppsala, Dorpat och Åbo under 1600-talet
- 2001
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The thesis focuses upon three universities in Sweden during its ‘Age of Greatness’, Uppsala, Dorpat and Åbo, with a special emphasis upon the universities academic jurisdiction. On the whole the universities saw State intervention as a useful tool in local conflicts with the surrounding town community but opposed that very same form of intervention when it tried to interfere in internal university matters. Only minor reforms however, in the face of often fierce State criticism as to the lack of discipline among the students, were instituted but left the universities’ prerogatives untouched.Relations with the local authorities were at times quite stormy and characterised by intense if brief disputes. The main areas of dispute between the universities and the local city authorities was the public peace and the rights to titles and ranking in Church during Mass. The results of the study show a marked difference between the three case studies in both areas which had important consequences for the relative success of social controls in general.The thesis has also investigated the internal relationship between the three universities themselves and the conclusion is that there was considerable difference between formal statures and actual practical jurisdiction. The relationship between professors and students were mutually dependent and beneficial. There were surprisingly few cases of them either verbally or physically abusing their lecturers. As for the students themselves the thesis looks at two vital aspects namely age and geographic distribution of origins. This illustrates the problem that some students were picked upon and punished or degraded because of their age or origins. Hardly any students responsible for these activities were however charged or punished for their actions.
|
|
| 7. |
- Gudmundson, Erik, 1976-
(författare)
-
Signal Processing for Spectroscopic Applications
- 2010
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Spectroscopic techniques allow for studies of materials and organisms on the atomic and molecular level. Examples of such techniques are nuclear magnetic resonance (NMR) spectroscopy—one of the principal techniques to obtain physical, chemical, electronic and structural information about molecules—and magnetic resonance imaging (MRI)—an important medical imaging technique for, e.g., visualization of the internal structure of the human body. The less well-known spectroscopic technique of nuclear quadrupole resonance (NQR) is related to NMR and MRI but with the difference that no external magnetic field is needed. NQR has found applications in, e.g., detection of explosives and narcotics. The first part of this thesis is focused on detection and identification of solid and liquid explosives using both NQR and NMR data. Methods allowing for uncertainties in the assumed signal amplitudes are proposed, as well as methods for estimation of model parameters that allow for non-uniform sampling of the data. The second part treats two medical applications. Firstly, new, fast methods for parameter estimation in MRI data are presented. MRI can be used for, e.g., the diagnosis of anomalies in the skin or in the brain. The presented methods allow for a significant decrease in computational complexity without loss in performance. Secondly, the estimation of blood flow velo-city using medical ultrasound scanners is addressed. Information about anomalies in the blood flow dynamics is an important tool for the diagnosis of, for example, stenosis and atherosclerosis. The presented methods make no assumption on the sampling schemes, allowing for duplex mode transmissions where B-mode images are interleaved with the Doppler emissions.
|
|
| 8. |
- Hamrin, Göran, 1970-
(författare)
-
Effective Domains and Admissible Domain Representations
- 2005
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This thesis consists of four papers in domain theory and a summary. The first two papers deal with the problem of defining effectivity for continuous cpos. The third and fourth paper present the new notion of an admissible domain representation, where a domain representation D of a space X is λ-admissible if, in principle, all other λ-based domain representations E of X can be reduced to X via a continuous function from E to D. In Paper I we define a cartesian closed category of effective bifinite domains. We also investigate the method of inducing effectivity onto continuous cpos via projection pairs, resulting in a cartesian closed category of projections of effective bifinite domains. In Paper II we introduce the notion of an almost algebraic basis for a continuous cpo, showing that there is a natural cartesian closed category of effective consistently complete continuous cpos with almost algebraic bases. We also generalise the notion of a complete set, used in Paper I to define the bifinite domains, and investigate what closure results that can be obtained. In Paper III we consider admissible domain representations of topological spaces. We present a characterisation theorem of exactly when a topological space has a λ-admissible and κ-based domain representation. We also show that there is a natural cartesian closed category of countably based and countably admissible domain representations. In Paper IV we consider admissible domain representations of convergence spaces, where a convergence space is a set X together with a convergence relation between nets on X and elements of X. We study in particular the new notion of weak κ-convergence spaces, which roughly means that the convergence relation satisfies a generalisation of the Kuratowski limit space axioms to cardinality κ. We show that the category of weak κ-convergence spaces is cartesian closed. We also show that the category of weak κ-convergence spaces that have a dense, λ-admissible, κ-continuous and α-based consistently complete domain representation is cartesian closed when α ≤ λ ≥ κ. As natural corollaries we obtain corresponding results for the associated category of weak convergence spaces.
|
|
| 9. |
- Isacs, Lina
(författare)
-
Deliberating value : On the theory and practice of valuation of nature from neoclassical to ecological economics
- 2021
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This thesis is about whether it is a good idea to place monetary value on nature, to remedy the fact that we treat it as having no particular value to us humans, although it clearly has. The thesis is based on five research papers that can be said to position themselves on opposite sides in the debate on monetisation of nature. The first two papers consider the basis of neoclassical environmental economics and apply the value theory and valuation methods from normative neoclassical welfare theory, on which monetisation of nature is based. The other three papers examine, with increasing degrees of criticism, this theory of value and especially its central assumption that value can be derived from people’s choices, or “revealed preferences”. The thesis itself is a “reflective story” about the journey I made as I learned to think about and understand neoclassical environmental economics in new ways. I reflect upon my work from a philosophy of science perspective, consider how for-granted-taken ideas from neoclassical economics affect environmental economic analysis and its conclusions, and examine the subject of value and valuation from what has become my new theoretical standpoint, ecological economics. It concerns meta-theoretical questions about ontology, that is, ideas in a research discipline about how things really are (what is), and epistemology, ideas about how researchers can provide relevant knowledge about reality. Such ideas are often taken for granted in neoclassical economic analysis and how they affect the analysis and its conclusions is not seldom unreflective. In the thesis, I move from explaining why neoclassical environmental economists advocate monetisation and pricing of nature as important solutions to environmental problems, to exemplifying how this turned out in research projects intended to serve as decision support in practice, and then to exploring and clarifying an alternative theory of value and valuation from ecological economics based on value pluralism and so-called deliberative valuation. In a concluding discussion, I point out that there are reasons to be sceptical about whether monetisation of nature is the right path to follow if we want to change our unsustainable relationship with nature and tackle the serious ecological crises we currently face. I show that monetisation of nature in practice requires a considerable amount of pragmatism, since the applied version of the theory deviates far from its idealised claims about the possibility to capture actual, total values. I also show that the descriptive (so-called positive) part of neoclassical theory and its normative part overlap in a way that makes it very difficult to speak of “objective” science in environmental economics. Instead, and despite strong recognition in the discipline that environmental problems are “market failures”, neoclassical theory has an ethical and ideological bias that favours individuals’ freedom of choice and market solutions, at the expense of collective decision-making and discussions about values that cannot be quantified. The important contribution of the thesis is that it clarifies the consequences of a central idea in the theory behind environmental economic analysis, namely the idea of values as commensurable, that is, measurable in one single unit. This idea links to the misleading conception of choices as “trade-offs”, where all choices are essentially viewed as the result of people’s constant exchange of costs and benefits within themselves in every choice they make, with the result that everything gets better (or at least not worse). Based on my research, I suggest that, in reality, people do not generally “make” trade-offs. If anything, people try to avoid them, especially when it comes to difficult choices, such as those concerning the true value of nature, because such choices involve moral conflicts between values that are incommensurable. As a basis for valuing transformational change, monetisation is therefore unsuitable, as it conceals rather than reveals the ethical dilemmas that are the very definition of sustainability problems and causes us to search for the efficient or so-called “optimal” solutions claimed possible in neoclassical theory and rhetoric, although such solutions do not exist. What we need instead is to represent public opinion in environmental decision-making in ways that do not conceal people’s actual moral considerations. Environmental valuation is political. It must be done together with others through reason-sensitive means, where people’s actual experiences of value conflicts – within us and between us – can be deliberated before making decisions. This makes decision-making more complex, but as an alternative to monetisation, this realism is not necessarily unrealistic. The fact that incommensurability is grounded in human experience means that the complexity of social and environmental decision-making has a real counterpart in conflicts within ourselves. One could see this as a potentiality, because we may have more confidence in people’s ability to recognise the relevance and necessity of less simplification and more complexity in decision-making. People need to “deliberate values” rather than “consuming” them and being expected to express all sorts of values through money.
|
|
| 10. |
|
|
