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.

Highly sensitive spectrophotometric methods for the determination of olanzapine

Highly sensitive and selective spectrophotometric methods (A and B) were developed for the determination of micro amounts of olanzapine (OLZ). Method A (direct method) is based on the oxidation of olanzapine with a known excess of iodine monochloride (ICl) in an acidic medium. Under the same condition, thymol blue was iodinated by unreacted ICl, and the absorbance of uniodinated thymol blue was measured at 536 nm. The decrease in ICl concentration is a measure of drug concentration. In method B (indirect method), oxidation of OLZ by a known excess of Ce(IV) in sulfuric acid medium followed by the reaction of unreacted Ce(IV) with leuco crystal violet (LCV) to crystal violet (CV), which is measured in an acetate buffer medium ( pH 4.9) at 580 nm. These methods obey the Beer's law in the concentration range of 0.2-1.6 µg mL-1 (method A) and 0.1-1.4 µg mL-1 (method B). The developed procedures have been successfully applied to the determination of OLZ in pure and in dosage forms. The results exhibit no interference from the presence of excipients. The reliability of the methods was established by parallel determination of OLZ against the reference method.

Microscopic simulation of pigment coating consolidation

The control of coating layer properties is becoming increasingly important as a result of an emerging demand for novel coated paper-based products and an increasing popularity of new coating application methods. The governing mechanisms of microstructure formation dynamics during consolidation and drying are nevertheless, still poorly understood. Some of the difficulties encountered by experimental methods can be overcome by the utilisation of numerical modelling and simulation-based studies of the consolidation process. The objective of this study was to improve the fundamental understanding of pigment coating consolidation and structure formation mechanisms taking place on the microscopic level. Furthermore, it is aimed to relate the impact of process and suspension properties to the microstructure of the coating layer. A mathematical model based on a modified Stokesian dynamics particle simulation technique was developed and applied in several studies of consolidation-related phenomena. The model includes particle-particle and particle-boundary hydrodynamics, colloidal interactions, Born repulsion, and a steric repulsion model. The Brownian motion and a free surface model were incorporated to enable the specific investigation of consolidation and drying. Filter cake stability was simulated in various particle systems, and subjected to a range of base substrate absorption rates and system temperatures. The stability of the filter cake was primarily affected by the absorption rate and size of particles. Temperature was also shown to have an influence. The consolidation of polydisperse systems, with varying wet coating thicknesses, was studied using imposed pilot trial and model-based drying conditions. The results show that drying methods have a clear influence on the microstructure development, on small particle distributions in the coating layer and also on the mobility of particles during consolidation. It is concluded that colloidal properties can significantly impact coating layer shrinkage as well as the internal solids concentration profile. Visualisations of particle system development in time and comparison of systems at different conditions are useful in illustrating coating layer structure formation mechanisms. The results aid in understanding the underlying mechanisms of pigment coating layer consolidation. Guidance is given regarding the relationship between coating process conditions and internal coating slurry properties and their effects on the microstructure of the coating.

Mathematical analysis of maximum power generated by photovoltaic systems and fitting curves for standard test conditions

The rural electrification is characterized by geographical dispersion of the population, low consumption, high investment by consumers and high cost. Moreover, solar radiation constitutes an inexhaustible source of energy and in its conversion into electricity photovoltaic panels are used. In this study, equations were adjusted to field conditions presented by the manufacturer for current and power of small photovoltaic systems. The mathematical analysis was performed on the photovoltaic rural system I-100 from ISOFOTON, with power 300 Wp, located at the Experimental Farm Lageado of FCA/UNESP. For the development of such equations, the circuitry of photovoltaic cells has been studied to apply iterative numerical methods for the determination of electrical parameters and possible errors in the appropriate equations in the literature to reality. Therefore, a simulation of a photovoltaic panel was proposed through mathematical equations that were adjusted according to the data of local radiation. The results have presented equations that provide real answers to the user and may assist in the design of these systems, once calculated that the maximum power limit ensures a supply of energy generated. This real sizing helps establishing the possible applications of solar energy to the rural producer and informing the real possibilities of generating electricity from the sun.

Estimation methods of reference evapotranspiration (ETo) for Uberlândia -MG

One approach to verify the adequacy of estimation methods of reference evapotranspiration is the comparison with the Penman-Monteith method, recommended by the United Nations of Food and Agriculture Organization - FAO, as the standard method for estimating ET0. This study aimed to compare methods for estimating ET0, Makkink (MK), Hargreaves (HG) and Solar Radiation (RS), with Penman-Monteith (PM). For this purpose, we used daily data of global solar radiation, air temperature, relative humidity and wind speed for the year 2010, obtained through the automatic meteorological station, with latitude 18° 91' 66" S, longitude 48° 25' 05" W and altitude of 869m, at the National Institute of Meteorology situated in the Campus of Federal University of Uberlandia - MG, Brazil. Analysis of results for the period were carried out in daily basis, using regression analysis and considering the linear model y = ax, where the dependent variable was the method of Penman-Monteith and the independent, the estimation of ET0 by evaluated methods. Methodology was used to check the influence of standard deviation of daily ET0 in comparison of methods. The evaluation indicated that methods of Solar Radiation and Penman-Monteith cannot be compared, yet the method of Hargreaves indicates the most efficient adjustment to estimate ETo.

Testate amoebae (thecamoebians) as indicators of aquatic mine impact

The environmental impacts of a single mine often remain local, but acidic and metal-rich acid mine drainage (AMD) from the waste materials may pose a serious threat to adjacent surface waters and their ecosystems. Testate amoebae (thecamoebian) analysis was used together with lake sediment geochemistry to study and evaluate the ecological effects of sulphidic metal mines on aquatic environments. Three different mines were included in the study: Luikonlahti Cu-mine in Kaavi, eastern Finland, Haveri Cu-Au mine in Ylöjärvi, southern Finland and Pyhäsalmi Zn-Cu-S mine in Pyhäjärvi, central Finland. Luikonlahti and Haveri are closed mines, but Pyhäsalmi is still operating. The sampling strategy was case specific, and planned to provide a representative sediment sample series to define natural background conditions, to detect spatial and temporal variations in mine impacts, to evaluate the possible recovery after the peak contamination, and to distinguish the effects of other environmental factors from the mining impacts. In the Haveri case, diatom analyses were performed alongside thecamoebian analysis to evaluate the similarities and differences between the two proxies. The results of the analyses were investigated with multivariate methods (direct and indirect ordinations, diversity and distance measure indices). Finally, the results of each case study were harmonized, pooled, and jointly analyzed to summarize the results for this dissertation. Geochemical results showed broadly similar temporal patterns in each case. Concentrations of ions in the pre-disturbance samples defined the natural baseline against which other results were compared. The beginning of the mining activities had only minor impacts on sediment geochemistry, mainly appearing as an increased clastic input into the lakes at Haveri and Pyhäsalmi. The active mining phase was followed by the metallic contamination and, subsequently, by the most recent change towards decreased but still elevated metal concentrations in the sediments. Because of the delay in the oxidation of waste material and formation of AMD, the most intense, but transient metal contamination phase occurred in the post-mining period at Luikonlahti and Haveri. At Pyhäsalmi, the highest metal contamination preceded effluent mitigation actions. Spatial gradients were observed besides the temporal evolution in both the pre-disturbance and mine-impacted samples from Luikonlahti and Pyhäsalmi. The geochemical gradients varied with distance from the main source of contaminants (dispersion and dilution) and with water depth (redox and pH). The spatial extent of the highest metal contamination associated with these mines remained rather limited. At Haveri, the metallic impact was widespread, with the upstream site in another lake basin found to be contaminated. Changes in thecamoebian assemblages corresponded well with the geochemical results. Despite some differences, the general features and ecological responses of the faunal assemblages were rather similar in each lake. Constantly abundant strains of Difflugia oblonga, Difflugia protaeiformis and centropyxids formed the core of these assemblages. Increasing proportions of Cucurbitella tricuspis towards the surface samples were found in all of the cases. The results affirmed the indicator value of some already known indicator forms, but such as C. tricuspis and higher nutrient levels, but also elicited possible new ones such as D. oblonga ‘spinosa’ and clayey substrate, high conductivity and/or alkalinity, D. protaeiformis ‘multicornis’ and pH, water hardness and the amount of clastic material and Centropyxis constricta ‘aerophila’ and high metal and S concentrations. In each case, eutrophication appeared to be the most important environmental factor, masking the effects of other variables. Faunal responses to high metal inputs in sediments remained minor, but were nevertheless detectable. Besides the trophic state of the lake, numerical methods suggested overall geochemical conditions (pH, redox) to be the most important factor at Luikonlahti, whereas the Haveri results showed the clearest connection between metals and amoebae. At Pyhäsalmi, the strongest relationships were found between Ca- and S-rich present loading, redox conditions and substrate composition. Sediment geochemistry and testate amoeba analysis proved to be a suitable combination of methods to detect and describe the aquatic mine impacts in each specific case, to evaluate recovery and to differentiate between the effects of different anthropogenic and natural environmental factors. It was also suggested that aquatic mine impacts can be significantly mitigated by careful design and after-care of the waste facilities, especially by reducing and preventing AMD. The case-specific approach is nevertheless necessary because of the unique characteristics of each mine and variations in the environmental background conditions.

Modelling of potentiometric ion sensors

Potentiometric ion sensors are a very important subgroup of electrochemical sensors, very attractive for practical applications due to their small size, portability, low-energy consumption, relatively low cost and not changing the sample composition. They are investigated by the researchers from many fields of science. The continuous development of this field creates the necessity for a detailed description of sensor response and the electrochemical processes important in the practical applications of ion sensors. The aim of this thesis is to present the existing models available for the description of potentiometric ion sensors as well as their applicability and limitations. This includes the description of the diffusion potential occurring at the reference electrodes. The wide range of existing models, from most idealised phase boundary models to most general models, including migration, is discussed. This work concentrates on the advanced modelling of ion sensors, namely the Nernst-Planck-Poisson (NPP) model, which is the most general of the presented models, therefore the most widely applicable. It allows the modelling of the transport processes occurring in ion sensors and generating the potentiometric response. Details of the solution of the NPP model (including the numerical methods used) are shown. The comparisons between NPP and the more idealized models are presented. The applicability of the model to describe the formation of diffusion potential in reference electrode, the lower detection limit of both ion-exchanger and neutral carrier electrodes and the effect of the complexation in the membrane are discussed. The model was applied for the description of both types of electrodes, i.e. with the inner filling solution and solidcontact electrodes. The NPP model allows the electrochemical methods other than potentiometry to be described. Application of this model in Electrochemical Impedance Spectroscopy is discussed and a possible use in chrono-potentiometry is indicated. By combining the NPP model with evolutionary algorithms, namely Hierarchical Genetic Strategy (HGS), a novel method allowing the facilitation of the design of ion sensors was created. It is described in detail in this thesis and its possible applications in the field of ion sensors are indicated. Finally, some interesting effects occurring in the ion sensors (i.e. overshot response and influence of anionic sites) as well as the possible applications of NPP in biochemistry are described.

Gas-solid two-phase flow in the riser of circulating fluidized beds: mathematical modelling and numerical simulation

A mathematical model is developed for gas-solids flows in circulating fluidized beds. An Eulerian formulation is followed based on the two-fluids model approach where both the fluid and the particulate phases are treated as a continuum. The physical modelling is discussed, including the formulation of boundary conditions and the description of the numerical methodology. Results of numerical simulation are presented and discussed. The model is validated through comparison to experiment, and simulation is performed to investigate the effects on the flow hydrodynamics of the solids viscosity.

Development of a process for the direct synthesis of hydrogen peroxide in a novel microstructured reactor

Microreactors have proven to be versatile tools for process intensification. Over recent decades, they have increasingly been used for product and process development in chemical industries. Enhanced heat and mass transfer in the reactors due to the extremely high surfacearea- to-volume ratio and interfacial area allow chemical processes to be operated at extreme conditions. Safety is improved by the small holdup volume of the reactors and effective control of pressure and temperature. Hydrogen peroxide is a powerful green oxidant that is used in a wide range of industries. Reduction and auto-oxidation of anthraquinones is currently the main process for hydrogen peroxide production. Direct synthesis is a green alternative and has potential for on-site production. However, there are two limitations: safety concerns because of the explosive gas mixture produced and low selectivity of the process. The aim of this thesis was to develop a process for direct synthesis of hydrogen peroxide utilizing microreactor technology. Experimental and numerical approaches were applied for development of the microreactor. Development of a novel microreactor was commenced by studying the hydrodynamics and mass transfer in prototype microreactor plates. The prototypes were designed and fabricated with the assistance of CFD modeling to optimize the shape and size of the microstructure. Empirical correlations for the mass transfer coefficient were derived. The pressure drop in micro T-mixers was investigated experimentally and numerically. Correlations describing the friction factor for different flow regimes were developed and predicted values were in good agreement with experimental results. Experimental studies were conducted to develop a highly active and selective catalyst with a proper form for the microreactor. Pd catalysts supported on activated carbon cloths were prepared by different treatments during the catalyst preparation. A variety of characterization methods were used for catalyst investigation. The surface chemistry of the support and the oxidation state of the metallic phase in the catalyst play important roles in catalyst activity and selectivity for the direct synthesis. The direct synthesis of hydrogen peroxide was investigated in a bench-scale continuous process using the novel microreactor developed. The microreactor was fabricated based on the hydrodynamic and mass transfer studies and provided a high interfacial area and high mass transfer coefficient. The catalysts were prepared under optimum treatment conditions. The direct synthesis was conducted at various conditions. The thesis represents a step towards a commercially viable direct synthesis. The focus is on the two main challenges: mitigating the safety problem by utilization of microprocess technology and improving the selectivity by catalyst development.

Evaluation of methods for estimating energy performance of biogas production

This study focused on identifying various system boundaries and evaluating methods of estimating energy performance of biogas production. First, the output-input ratio method used for evaluating energy performance from the system boundaries was reviewed. Secondly, ways to assess the efficiency of biogas use and parasitic energy demand were investigated. Thirdly, an approach for comparing biogas production to other energy production methods was evaluated. Data from an existing biogas plant, located in Finland, was used for the evaluation of the methods. The results indicate that calculating and comparing the output-input ratios (Rpr1, Rpr2, Rut, Rpl and Rsy) can be used in evaluating the performance of biogas production system. In addition, the parasitic energy demand calculations (w) and the efficiency of utilizing produced biogas (η) provide detailed information on energy performance of the biogas plant. Furthermore, Rf and energy output in relation to total solid mass of feedstock (FO/TS) are useful in comparing biogas production with other energy recovery technologies. As a conclusion it is essential for the comparability of biogas plants that their energy performance would be calculated in a more consistent manner in the future.

Adaptive Markov chain Monte Carlo and Bayesian filtering for state space models

This thesis is concerned with the state and parameter estimation in state space models. The estimation of states and parameters is an important task when mathematical modeling is applied to many different application areas such as the global positioning systems, target tracking, navigation, brain imaging, spread of infectious diseases, biological processes, telecommunications, audio signal processing, stochastic optimal control, machine learning, and physical systems. In Bayesian settings, the estimation of states or parameters amounts to computation of the posterior probability density function. Except for a very restricted number of models, it is impossible to compute this density function in a closed form. Hence, we need approximation methods. A state estimation problem involves estimating the states (latent variables) that are not directly observed in the output of the system. In this thesis, we use the Kalman filter, extended Kalman filter, Gauss–Hermite filters, and particle filters to estimate the states based on available measurements. Among these filters, particle filters are numerical methods for approximating the filtering distributions of non-linear non-Gaussian state space models via Monte Carlo. The performance of a particle filter heavily depends on the chosen importance distribution. For instance, inappropriate choice of the importance distribution can lead to the failure of convergence of the particle filter algorithm. In this thesis, we analyze the theoretical Lᵖ particle filter convergence with general importance distributions, where p ≥2 is an integer. A parameter estimation problem is considered with inferring the model parameters from measurements. For high-dimensional complex models, estimation of parameters can be done by Markov chain Monte Carlo (MCMC) methods. In its operation, the MCMC method requires the unnormalized posterior distribution of the parameters and a proposal distribution. In this thesis, we show how the posterior density function of the parameters of a state space model can be computed by filtering based methods, where the states are integrated out. This type of computation is then applied to estimate parameters of stochastic differential equations. Furthermore, we compute the partial derivatives of the log-posterior density function and use the hybrid Monte Carlo and scaled conjugate gradient methods to infer the parameters of stochastic differential equations. The computational efficiency of MCMC methods is highly depend on the chosen proposal distribution. A commonly used proposal distribution is Gaussian. In this kind of proposal, the covariance matrix must be well tuned. To tune it, adaptive MCMC methods can be used. In this thesis, we propose a new way of updating the covariance matrix using the variational Bayesian adaptive Kalman filter algorithm.

Thermal reactions of the major hydrocarbon components of biomass gasification gas

Gasification of biomass is an efficient method process to produce liquid fuels, heat and electricity. It is interesting especially for the Nordic countries, where raw material for the processes is readily available. The thermal reactions of light hydrocarbons are a major challenge for industrial applications. At elevated temperatures, light hydrocarbons react spontaneously to form higher molecular weight compounds. In this thesis, this phenomenon was studied by literature survey, experimental work and modeling effort. The literature survey revealed that the change in tar composition is likely caused by the kinetic entropy. The role of the surface material is deemed to be an important factor in the reactivity of the system. The experimental results were in accordance with previous publications on the subject. The novelty of the experimental work lies in the used time interval for measurements combined with an industrially relevant temperature interval. The aspects which are covered in the modeling include screening of possible numerical approaches, testing of optimization methods and kinetic modelling. No significant numerical issues were observed, so the used calculation routines are adequate for the task. Evolutionary algorithms gave a better performance combined with better fit than the conventional iterative methods such as Simplex and Levenberg-Marquardt methods. Three models were fitted on experimental data. The LLNL model was used as a reference model to which two other models were compared. A compact model which included all the observed species was developed. The parameter estimation performed on that model gave slightly impaired fit to experimental data than LLNL model, but the difference was barely significant. The third tested model concentrated on the decomposition of hydrocarbons and included a theoretical description of the formation of carbon layer on the reactor walls. The fit to experimental data was extremely good. Based on the simulation results and literature findings, it is likely that the surface coverage of carbonaceous deposits is a major factor in thermal reactions.

Advanced analysis and design methods for preparative chromatographic separation processes

Preparative liquid chromatography is one of the most selective separation techniques in the fine chemical, pharmaceutical, and food industries. Several process concepts have been developed and applied for improving the performance of classical batch chromatography. The most powerful approaches include various single-column recycling schemes, counter-current and cross-current multi-column setups, and hybrid processes where chromatography is coupled with other unit operations such as crystallization, chemical reactor, and/or solvent removal unit. To fully utilize the potential of stand-alone and integrated chromatographic processes, efficient methods for selecting the best process alternative as well as optimal operating conditions are needed. In this thesis, a unified method is developed for analysis and design of the following singlecolumn fixed bed processes and corresponding cross-current schemes: (1) batch chromatography, (2) batch chromatography with an integrated solvent removal unit, (3) mixed-recycle steady state recycling chromatography (SSR), and (4) mixed-recycle steady state recycling chromatography with solvent removal from fresh feed, recycle fraction, or column feed (SSR–SR). The method is based on the equilibrium theory of chromatography with an assumption of negligible mass transfer resistance and axial dispersion. The design criteria are given in general, dimensionless form that is formally analogous to that applied widely in the so called triangle theory of counter-current multi-column chromatography. Analytical design equations are derived for binary systems that follow competitive Langmuir adsorption isotherm model. For this purpose, the existing analytic solution of the ideal model of chromatography for binary Langmuir mixtures is completed by deriving missing explicit equations for the height and location of the pure first component shock in the case of a small feed pulse. It is thus shown that the entire chromatographic cycle at the column outlet can be expressed in closed-form. The developed design method allows predicting the feasible range of operating parameters that lead to desired product purities. It can be applied for the calculation of first estimates of optimal operating conditions, the analysis of process robustness, and the early-stage evaluation of different process alternatives. The design method is utilized to analyse the possibility to enhance the performance of conventional SSR chromatography by integrating it with a solvent removal unit. It is shown that the amount of fresh feed processed during a chromatographic cycle and thus the productivity of SSR process can be improved by removing solvent. The maximum solvent removal capacity depends on the location of the solvent removal unit and the physical solvent removal constraints, such as solubility, viscosity, and/or osmotic pressure limits. Usually, the most flexible option is to remove solvent from the column feed. Applicability of the equilibrium design for real, non-ideal separation problems is evaluated by means of numerical simulations. Due to assumption of infinite column efficiency, the developed design method is most applicable for high performance systems where thermodynamic effects are predominant, while significant deviations are observed under highly non-ideal conditions. The findings based on the equilibrium theory are applied to develop a shortcut approach for the design of chromatographic separation processes under strongly non-ideal conditions with significant dispersive effects. The method is based on a simple procedure applied to a single conventional chromatogram. Applicability of the approach for the design of batch and counter-current simulated moving bed processes is evaluated with case studies. It is shown that the shortcut approach works the better the higher the column efficiency and the lower the purity constraints are.

Comparison of the quantitative competitive and semiquantitative RT-PCR methods for the determination of interferon-gamma mRNA levels in AIDS-free HIV-infected individuals

IFN-gamma mRNA expression was evaluated in nonstimulated peripheral blood mononuclear cells (PBMC) of HIV-infected and seronegative individuals using quantitative competitive and semiquantitative RT-PCR and the sensitivity of these methods was compared. A significant correlation was found between quantitative competitive and semiquantitative RT-PCR in samples of both HIV-seronegative (P = 0.004) and HIV-infected individuals (P = 0.0004). PBMC from HIV-infected individuals presented a remarkable increase of IFN-gamma mRNA expression, as determined by both types of RT-PCR methods. Semiquantitative RT-PCR even without an internal standard is also acceptable for measuring cytokine mRNA expression, but less reliable if small amounts are quantified. Moreover, we found that increased IFN-gammamRNA expression is independent of CD4+ cell count in AIDS-free HIV-infected patients.