Learning Preferences with Kernel-Based Methods

Learning of preference relations has recently received significant attention in machine learning community. It is closely related to the classification and regression analysis and can be reduced to these tasks. However, preference learning involves prediction of ordering of the data points rather than prediction of a single numerical value as in case of regression or a class label as in case of classification. Therefore, studying preference relations within a separate framework facilitates not only better theoretical understanding of the problem, but also motivates development of the efficient algorithms for the task. Preference learning has many applications in domains such as information retrieval, bioinformatics, natural language processing, etc. For example, algorithms that learn to rank are frequently used in search engines for ordering documents retrieved by the query. Preference learning methods have been also applied to collaborative filtering problems for predicting individual customer choices from the vast amount of user generated feedback. In this thesis we propose several algorithms for learning preference relations. These algorithms stem from well founded and robust class of regularized least-squares methods and have many attractive computational properties. In order to improve the performance of our methods, we introduce several non-linear kernel functions. Thus, contribution of this thesis is twofold: kernel functions for structured data that are used to take advantage of various non-vectorial data representations and the preference learning algorithms that are suitable for different tasks, namely efficient learning of preference relations, learning with large amount of training data, and semi-supervised preference learning. Proposed kernel-based algorithms and kernels are applied to the parse ranking task in natural language processing, document ranking in information retrieval, and remote homology detection in bioinformatics domain. Training of kernel-based ranking algorithms can be infeasible when the size of the training set is large. This problem is addressed by proposing a preference learning algorithm whose computation complexity scales linearly with the number of training data points. We also introduce sparse approximation of the algorithm that can be efficiently trained with large amount of data. For situations when small amount of labeled data but a large amount of unlabeled data is available, we propose a co-regularized preference learning algorithm. To conclude, the methods presented in this thesis address not only the problem of the efficient training of the algorithms but also fast regularization parameter selection, multiple output prediction, and cross-validation. Furthermore, proposed algorithms lead to notably better performance in many preference learning tasks considered.

Factors predicting mortality in type 2 diabetes. With special reference to physical exercise, blood pressure, proteinuria, inflammation and P wave duration

Background: Type 2 diabetes patients have a 2-4 fold risk of cardiovascular disease (CVD) compared to the general population. In type 2 diabetes, several CVD risk factors have been identified, including obesity, hypertension, hyperglycemia, proteinuria, sedentary lifestyle and dyslipidemia. Although much of the excess CVD risk can be attributed to these risk factors, a significant proportion is still unknown. Aims: To assess in middle-aged type 2 diabetic subjects the joint relations of several conventional and non-conventional CVD risk factors with respect to cardiovascular and total mortality. Subjects and methods: This thesis is part of a large prospective, population based East-West type 2 diabetes study that was launched in 1982-1984. It includes 1,059 middle-aged (45-64 years old) participants. At baseline, a thorough clinical examination and laboratory measurements were performed and an ECG was recorded. The latest follow-up study was performed 18 years later in January 2001 (when the subjects were 63-81 years old). The study endpoints were total mortality and mortality due to CVD, coronary heart disease (CHD) and stroke. Results: Physically more active patients had significantly reduced total, CVD and CHD mortality independent of high-sensitivity C-reactive protein (hs-CRP) levels unless proteinuria was present. Among physically active patients with a hs-CRP level >3 mg/L, the prognosis of CVD mortality was similar to patients with hs-CRP levels ≤3 mg/L. The worst prognosis was among physically inactive patients with hs-CRP levels >3 mg/L. Physically active patients with proteinuria had significantly increased total and CVD mortality by multivariate analyses. After adjustment for confounding factors, patients with proteinuria and a systolic BP <130 mmHg had a significant increase in total and CVD mortality compared to those with a systolic BP between 130 and 160 mmHg. The prognosis was similar in patients with a systolic BP <130 mmHg and ≥160 mmHg. Among patients without proteinuria, a systolic BP <130 mmHg was associated with a non-significant reduction in mortality. A P wave duration ≥114 ms was associated with a 2.5-fold increase in stroke mortality among patients with prevalent CHD or claudication. This finding persisted in multivariable analyses. Among patients with no comorbidities, there was no relationship between P wave duration and stroke mortality. Conclusions: Physical activity reduces total and CVD mortality in patients with type 2 diabetes without proteinuria or with elevated levels of hs-CRP, suggesting that the anti-inflammatory effect of physical activity can counteract increased CVD morbidity and mortality associated with a high CRP level. In patients with proteinuria the protective effect was not, however, present. Among patients with proteinuria, systolic BP <130 mmHg may increase mortality due to CVD. These results demonstrate the importance of early intervention to prevent CVD and to control all-cause mortality among patients with type 2 diabetes. The presence of proteinuria should be taken into account when defining the target systolic BP level for prevention of CVD deaths. A prolongation of the duration of the P wave was associated with increased stroke mortality among high-risk patients with type 2 diabetes. P wave duration is easy to measure and merits further examination to evaluate its importance for estimation of the risk of stroke among patients with type 2 diabetes.

Thyroid function in the elderly and sex hormones in elderly men: reference intervals and associations with health

Aims: The aims were to create clinically feasible reference intervals for thyroidstimulating hormone (TSH) and free thyroxine (FT4) and to analyze associations between thyroid function and self-rated health, neuropsychiatric symptoms, depression and dementia in the elderly. The second aim was also to establish reference intervals for sex hormones and to analyze associations between sex hormone levels and self-rated health, symptoms, depression and dementia in elderly men. Subjects and methods: The study population comprised 1252 subjects aged 65 years or over, living in the municipality of Lieto, south-western Finland. Self-rated health, life satisfaction, symptoms, depression, and dementia were assessed with specific questions, clinical examination and tools such as the Zung Self-report Depression Scale and the Mini-Mental State Examination. Independent variables were dichotomized, and associations of these variables with TSH, FT4 or sex hormone levels were assessed. Levels of TSH and FT4 in thyroid disease–free women and women treated with thyroxine were also compared. Results: Elevated concentrations of thyroid peroxidase antibodies (TPOAb) or thyroglobulin antibodies (TgAb) were found to have a marked effect on the upper reference limit for TSH among women, who were thyroid antibody positive more higher than suggested in several recent guidelines. After age adjustment, there were no associations between TSH levels and self-rated health, life satisfaction, or most neuropsychiatric symptoms in the thyroid disease-free population. Although women with thyroxine treatment for primary hypothyroidism had far higher TSH levels than thyroid disease-free women, there were no differences between thyroid-disease free women and women with stable thyroxine treatment regarding self-rated health, life satisfaction or symptoms. Age had a significant positive association with luteinizing hormone (LH), follicle 2 practice, one range in men aged 65 years or over can be used for T, E2 and FSH measured with the AutoDelfia method, but two separate reference intervals should be used for fT, LH and SHBG. After adjustment for age, higher levels of T and fT were associated with better self-rated health (SRH) in the reference population. After adjustment for age and body mass index (BMI), there were no associations between sex hormone concentrations and self-rated health, life satisfaction or most symptoms in concentration. Conclusion: Age-specific reference intervals were derived for thyroid function and sex hormones based on comprehensive data from a community-dwelling population with a high participation rate. The results do not support the need to decrease the upper reference limit for TSH or to lower the optimal TSH target in levothyroxine treatment in older adults, as recommended in recent guidelines. Older age or being overweight symptoms among elderly men. The associations of single symptoms with T levels were inconsistent among elderly men, although the association of low T level with diagnosed depression might be clinically significant.

Metabolic syndrome – early cardio-metabolic, vascular and hepatic changes

Background: Metabolic syndrome (MetS) is a combination of several cardio-metabolic risk factors including obesity, hyperglycemia, hypertension and dyslipidemia. MetS has been associated with increased levels of apolipoprotein B (apoB) and low-density lipoprotein oxidation (OxLDL) and with an increased risk of cardiovascular disease and non-alcoholic fatty liver disease. Aims: To establish the relation of apoB and OxLDL with the MetS development and to determine the status of MetS as a risk factor for adverse liver changes and for subclinical atherosclerosis. Subjects and Methods: The present thesis is part of the two large scale population-based, prospective, observational studies. Cardiovascular Risk in Young Finns study was launched in 1980 including 3,596 subjects aged 3-18 years. Thereafter follow-up studies have been conducted regularly. In the latest follow-ups that were performed in 2001 (N=2,283) and 2007 (N=2,204), non-invasive ultrasound studies were introduced to the study protocol to measure subclinical atherosclerosis i.e. carotid intima-media thickness (IMT), carotid artery distensibility (Cdist) and brachial flow-mediated dilatation (FMD). Alanine-aminotransferase (ALT) and gammaglutamyltransferase (GGT) were measured in 2007 to assess liver function. The Bogalusa Heart Study is a long-term epidemiologic study of cardiovascular risk factors launched in 1972 in a biracial community of Bogalusa, Louisiana, USA. Total of 374 youths (aged 9-18 years at baseline in 1984-88) who underwent non-invasive ultrasound studies of the carotid artery as adults, were included in the analyses of the present thesis. Results: The odds ratios (95% confidence intervals) for MetS incidence during a 6-year follow-up by quartiles of apoB were 2.0(1.0-3.8) for the second quartile, 3.1(1.7-5.7) for the third quartile and 4.2(2.3-7.6) for the fourth quartile. OxLDL was not independently associated with incident MetS. Youth (aged 9-18 years) with MetS or with high body mass index were at 2-3 times the risk of having MetS, high IMT, and type 2 diabetes 24-years later as adults. IMT increased 79±7μm (mean±SEM) in subjects with MetS and 42±2μm in subjects without the MetS (P<0.0001) during 6- years. Subjects who lost the MetS diagnosis during 6-year follow-up had reduced IMT progression compared to persistent MetS group (0.036±0.005vs.0.079±0.010 mm, P=0.001) and reduced Cdist change compared to incident MetS group (-0.12±0.05vs.-0.38±0.10 %/mmHg, P=0.03) over 6-year follow-up. MetS predicted elevated ALT (β±SEM=0.380±0.052, P<0.0001 in men and 0.160±0.052, P=0.002 in women) and GGT (β±SEM=0.240±0.058, P<0.0001 in men and 0.262±0.053, P<0.0001 in women) levels after 6-years. Conclusions: These findings suggest that apoB may give additional information on early metabolic disturbances predisposing MetS. MetS may be used to identify individuals at increased risk of developing atherosclerosis and non-alcoholic liver disease. However, recovery from the MetS may have positive effects on liver and vascular properties.

Characterisation of Laser Beam and Paper Material Interaction

It is known already from 1970´s that laser beam is suitable for processing paper materials. In this thesis, term paper materials mean all wood-fibre based materials, like dried pulp, copy paper, newspaper, cardboard, corrugated board, tissue paper etc. Accordingly, laser processing in this thesis means all laser treatments resulting material removal, like cutting, partial cutting, marking, creasing, perforation etc. that can be used to process paper materials. Laser technology provides many advantages for processing of paper materials: non-contact method, freedom of processing geometry, reliable technology for non-stop production etc. Especially packaging industry is very promising area for laser processing applications. However, there are only few industrial laser processing applications worldwide even in beginning of 2010´s. One reason for small-scale use of lasers in paper material manufacturing is that there is a shortage of published research and scientific articles. Another problem, restraining the use of laser for processing of paper materials, is colouration of paper material i.e. the yellowish and/or greyish colour of cut edge appearing during cutting or after cutting. These are the main reasons for selecting the topic of this thesis to concern characterization of interaction of laser beam and paper materials. This study was carried out in Laboratory of Laser Processing at Lappeenranta University of Technology (Finland). Laser equipment used in this study was TRUMPF TLF 2700 carbon dioxide laser that produces a beam with wavelength of 10.6 μm with power range of 190-2500 W (laser power on work piece). Study of laser beam and paper material interaction was carried out by treating dried kraft pulp (grammage of 67 g m-2) with different laser power levels, focal plane postion settings and interaction times. Interaction between laser beam and dried kraft pulp was detected with different monitoring devices, i.e. spectrometer, pyrometer and active illumination imaging system. This way it was possible to create an input and output parameter diagram and to study the effects of input and output parameters in this thesis. When interaction phenomena are understood also process development can be carried out and even new innovations developed. Fulfilling the lack of information on interaction phenomena can assist in the way of lasers for wider use of technology in paper making and converting industry. It was concluded in this thesis that interaction of laser beam and paper material has two mechanisms that are dependent on focal plane position range. Assumed interaction mechanism B appears in range of average focal plane position of 3.4 mm and 2.4 mm and assumed interaction mechanism A in range of average focal plane position of 0.4 mm and -0.6 mm both in used experimental set up. Focal plane position 1.4 mm represents midzone of these two mechanisms. Holes during laser beam and paper material interaction are formed gradually: first small hole is formed to interaction area in the centre of laser beam cross-section and after that, as function of interaction time, hole expands, until interaction between laser beam and dried kraft pulp is ended. By the image analysis it can be seen that in beginning of laser beam and dried kraft pulp material interaction small holes off very good quality are formed. It is obvious that black colour and heat affected zone appear as function of interaction time. This reveals that there still are different interaction phases within interaction mechanisms A and B. These interaction phases appear as function of time and also as function of peak intensity of laser beam. Limit peak intensity is the value that divides interaction mechanism A and B from one-phase interaction into dual-phase interaction. So all peak intensity values under limit peak intensity belong to MAOM (interaction mechanism A one-phase mode) or to MBOM (interaction mechanism B onephase mode) and values over that belong to MADM (interaction mechanism A dual-phase mode) or to MBDM (interaction mechanism B dual-phase mode). Decomposition process of cellulose is evolution of hydrocarbons when temperature is between 380- 500°C. This means that long cellulose molecule is split into smaller volatile hydrocarbons in this temperature range. As temperature increases, decomposition process of cellulose molecule changes. In range of 700-900°C, cellulose molecule is mainly decomposed into H2 gas; this is why this range is called evolution of hydrogen. Interaction in this range starts (as in range of MAOM and MBOM), when a small good quality hole is formed. This is due to “direct evaporation” of pulp via decomposition process of evolution of hydrogen. And this can be seen can be seen in spectrometer as high intensity peak of yellow light (in range of 588-589 nm) which refers to temperature of ~1750ºC. Pyrometer does not detect this high intensity peak since it is not able to detect physical phase change from solid kraft pulp to gaseous compounds. As interaction time between laser beam and dried kraft pulp continues, hypothesis is that three auto ignition processes occurs. Auto ignition of substance is the lowest temperature in which it will spontaneously ignite in a normal atmosphere without an external source of ignition, such as a flame or spark. Three auto ignition processes appears in range of MADM and MBDM, namely: 1. temperature of auto ignition of hydrogen atom (H2) is 500ºC, 2. temperature of auto ignition of carbon monoxide molecule (CO) is 609ºC and 3. temperature of auto ignition of carbon atom (C) is 700ºC. These three auto ignition processes leads to formation of plasma plume which has strong emission of radiation in range of visible light. Formation of this plasma plume can be seen as increase of intensity in wavelength range of ~475-652 nm. Pyrometer shows maximum temperature just after this ignition. This plasma plume is assumed to scatter laser beam so that it interacts with larger area of dried kraft pulp than what is actual area of beam cross-section. This assumed scattering reduces also peak intensity. So result shows that assumably scattered light with low peak intensity is interacting with large area of hole edges and due to low peak intensity this interaction happens in low temperature. So interaction between laser beam and dried kraft pulp turns from evolution of hydrogen to evolution of hydrocarbons. This leads to black colour of hole edges.

Testing Platform Development for Large Scale Solar Integration

Currently, a high penetration level of Distributed Generations (DGs) has been observed in the Danish distribution systems, and even more DGs are foreseen to be present in the upcoming years. How to utilize them for maintaining the security of the power supply under the emergency situations, has been of great interest for study. This master project is intended to develop a control architecture for studying purposes of distribution systems with large scale integration of solar power. As part of the EcoGrid EU Smart Grid project, it focuses on the system modelling and simulation of a Danish representative LV network located in Bornholm island. Regarding the control architecture, two types of reactive control techniques are implemented and compare. In addition, a network voltage control based on a tap changer transformer is tested. The optimized results after applying a genetic algorithm to five typical Danish domestic loads are lower power losses and voltage deviation using Q(U) control, specially with large consumptions. Finally, a communication and information exchange system is developed with the objective of regulating the reactive power and thereby, the network voltage remotely and real-time. Validation test of the simulated parameters are performed as well.

Oxidized LDL and physical fitness in healthy young men: associations with body composition, smoking, metabolic syndrome and androgen status

Background: In the past, oxidized low density lipoprotein (ox-LDL) has been associated with an unbeneficial lipid profile. This atherogenic lipid profile increases the risk of atherosclerotic cardiovascular diseases. Physical fitness has substantial effect on serum lipoprotein concentration as well as body composition and humoral responses, however interrelationships between ox-LDL and physical fitness have not been widely examined in a nationally representative sample. Aims: This thesis evaluates how cardiorespiratory and muscular fitness associate with ox-LDL lipids and how the other known risk factors of atherosclerosis might alter these associations. Subjects and Methods: The study cohort consisted of 846 healthy young males (mean age 25.1, SD 4.6) who were gathered by voluntary nationwide recruitment. Each participant conducted a series of physical fitness tests (cardiorespiratory and muscular fitness) and answered a detailed questionnaire that included lifestyle habits (i.e. smoking and leisuretime physical activity). Venous blood samples including ox-LDL and serum lipids were also collected. Results: Higher levels of ox-LDL were found in overweight and obese men, however, high cardiorespiratory fitness seemed to protect the overweight from high levels of ox-LDL. Young men who smoked and had poor cardiorespiratory or muscular fitness possessed a higher concentration of ox-LDL lipids when compared to comparable levels of cardiorespiratory or muscular fitness non-smoking young men. Metabolic syndrome was associated with increased levels of ox-LDL and high levels of ox-LDL combined with poor cardiorespiratory and abdominal muscle fitness seems to predict metabolic syndrome in young men. Also, participants with poor cardiorespiratory fitness and low levels of testosterone had higher levels of ox-LDL when compared to participants with high cardiorespiratory fitness / low testosterone as well as those with poor cardiorespiratory fitness / high testosterone. Conclusions: Good cardiorespiratory and muscular fitness protects young men from increased levels of ox-LDL lipids. This association was discovered in young men who were categorized as being overweight, smokers, metabolic syndrome or with low levels of testosterone. Being fit seems to prevent higher levels of ox-LDL, even in young healthy

A Model-Based Development and Verification Framework for Distributed System-on-Chip Architecture

The capabilities and thus, design complexity of VLSI-based embedded systems have increased tremendously in recent years, riding the wave of Moore’s law. The time-to-market requirements are also shrinking, imposing challenges to the designers, which in turn, seek to adopt new design methods to increase their productivity. As an answer to these new pressures, modern day systems have moved towards on-chip multiprocessing technologies. New architectures have emerged in on-chip multiprocessing in order to utilize the tremendous advances of fabrication technology. Platform-based design is a possible solution in addressing these challenges. The principle behind the approach is to separate the functionality of an application from the organization and communication architecture of hardware platform at several levels of abstraction. The existing design methodologies pertaining to platform-based design approach don’t provide full automation at every level of the design processes, and sometimes, the co-design of platform-based systems lead to sub-optimal systems. In addition, the design productivity gap in multiprocessor systems remain a key challenge due to existing design methodologies. This thesis addresses the aforementioned challenges and discusses the creation of a development framework for a platform-based system design, in the context of the SegBus platform - a distributed communication architecture. This research aims to provide automated procedures for platform design and application mapping. Structural verification support is also featured thus ensuring correct-by-design platforms. The solution is based on a model-based process. Both the platform and the application are modeled using the Unified Modeling Language. This thesis develops a Domain Specific Language to support platform modeling based on a corresponding UML profile. Object Constraint Language constraints are used to support structurally correct platform construction. An emulator is thus introduced to allow as much as possible accurate performance estimation of the solution, at high abstraction levels. VHDL code is automatically generated, in the form of “snippets” to be employed in the arbiter modules of the platform, as required by the application. The resulting framework is applied in building an actual design solution for an MP3 stereo audio decoder application.