This thesis compares hand-designed features with features learned by feature learning methods in video classification. The features learned by Principal Component Analysis whitening, Independent subspace analysis and Sparse Autoencoders were tested in a standard Bag of Visual Word classification paradigm replacing hand-designed features (e.g. SIFT, HOG, HOF). The classification performance was measured on Human Motion DataBase and YouTube Action Data Set. Learned features showed better performance than the hand-desined features. The combination of hand-designed features and learned features by Multiple Kernel Learning method showed even better performance, including cases when hand-designed features and learned features achieved not so good performance separately.

This thesis deals with the increasingly developing field of biometric systems which is the identification of faces. The thesis deals with the possibilities of face localization in pictures and their normalization, which is necessary due to external influences and the influence of different scanning techniques. It describes various techniques of localization of dominant features of the face such as eyes, mouth or nose. Not least, it describes different approaches to the identification of faces. Furthermore a it deals with an implementation of the Dominant Face Features Recognition application, which demonstrates chosen methods for localization of the dominant features (Hough Transform for Circles, localization of mouth using the location of the eyes) and for identification of a face (Linear Discriminant Analysis, Kernel Discriminant Analysis). The last part of the thesis contains a summary of achieved results and a discussion.

Les processus biologiques sont nombreux et leurs interactions complexes. Les mesures de cesphénomènes génèrent des biomarqueurs multiples. Ainsi, l’épidémiologie doit évoluer dans cecontexte de données complexes et de nature multidimensionnelle. Les maladies du systèmeimmunitaire et les troubles immunologiques qui leur sont associés constituent un bon exemplede pathologies où les questions clinico-épidémiologiques sont de plus en plus complexes,nécessitant des méthodes biostatistiques et épidémiologiques adaptées. Dans cette thèsed’Université, des méthodes permettant de prendre en compte les difficultés méthodologiquesgénérées par les données d’immunologie sont présentées autour de trois applicationscliniques. Notre approche consiste en l’utilisation de méthodes intégratives qui prennent encompte l’ensemble des mesures concernant une pathologie donnée. Nous montrons l’intérêtde l’analyse en composantes principales et de la classification hiérarchique ascendante pourrésumer et extraire l’information de données multiples de cytométrie en flux et celui desmodèles d’équations structurelles pour l’étude de relations complexes entre variables lors deprocessus multifactoriels. Enfin, via l’exemple d’un modèle de reconstitution immunitaireasymptotique utilisant une fonction exponentielle, nous illustrons l’importance de s’appuyersur la structure même des données et sur la compréhension des mécanismes biologiques quisous-tendent la variabilité de ces données dans la réflexion qui concourt au choix d’un modèlestatistique. Les méthodes et la réflexion proposées dans cette thèse sont transposables àd’autres domaines d’application avec des données multiples complexes. / Numerous biological processes with potentially complex interactions exist. Measurements ofthese processes allow to produce multiple biomarkers. Thus, there is a need for epidemiologyto evolve within the context of complex and multidimensional data. Immune system diseasesand associated immune disorders are an example of a field where clinical and epidemiologicalissues are increasingly complex, requiring appropriate statistical and epidemiologicalmethods. In this thesis, methods taking into account methodological difficulties generated byimmunology data are presented through three motivating examples. The general paradigm ofour approach is to take into account all measurements on a given pathology using integrativemethods. We propose principal component analysis and hierarchical clustering to summarizemultidimensional cytometry data and structural equation modelling for dealing with complexrelationships between variables in multifactorial processes. Then, through the example of anasymptotic model of immune reconstitution using an exponential function, we illustrate theimportance about the data’s structure and the biological mechanisms underlying its variabilitywhen building a mathematical model. The methods and the thinking advocated in this thesisare transposable to other research domains with complex data.

This thesis is carried out in collaboration with a private company, DANSiC AB This study is an extension of a research work started by DANSiC AB in 2019 to classify a source. This study is about classifying a source into two classes with the sensitivity of one source higher than the other as one source has greater importance. The data provided for this thesis is based on sensor measurements on different temperature cycles. The data is high-dimensional and is expected to have a drift in measurements. Principal component analysis (PCA) is used for dimensionality reduction. “Differential”, “Relative” and “Fractional” drift compensation techniques are used for compensating the drift in data. A comparative study was performed using three different classification algorithms, which are “Linear Discriminant Analysis (LDA)”, “Naive Bayes classifier (NB)” and “Random forest (RF)”. The highest accuracy achieved is 59%,Random forest is observed to perform better than the other classifiers. / <p>This work is done with DANSiC AB in collaboration with Linkoping University.</p>

Design space exploration (DSE) is used to improve and understand engineering designs. Such designs must meet objectives and structural requirements. Design improvement is non-trivial and requires new DSE methods. Turbomachinery manufacturers must continue to improve existing engines to keep up with global demand. Two challenges of turbomachinery DSE are: the time required to evaluate designs, and knowing which designs to evaluate. This research addressed these challenges by developing novel surrogate and principal component analysis (PCA) based DSE methods. Node and PCA-based surrogates were created to allow faster DSE of turbomachinery blades. The surrogates provided static stress estimation within 10% error. Surrogate error was related to the number of sampled finite element (FE) models used to train the surrogate and the variables used to change the designs. Surrogates were able to provide structural evaluations three to five orders of magnitude faster than FEA evaluations. The PCA-based surrogates were then used to create a PCA-based design workflow to help designers know which designs to evaluate. The workflow used either two-point correlation or stress and geometry coupling to relate the design variables to principal component (PC) scores. These scores were projections of the FE models onto the PCs obtained from PCA. Analysis showed that this workflow could be used in DSE to better explore and improve designs. The surrogate methods were then applied to vibratory stress. A computationally simplified analysis workflow was developed to allow for enough fluid and structural analyses to create a surrogate model. The simplified analysis workflow introduced 10% error but decreased the computational cost by 90%. The surrogate methods could not directly be applied to emulation of vibration due to the large spikes which occur near resonance. A novel, indirect emulation method was developed to better estimate vibratory responses Surrogates were used to estimate the inputs to calculate the vibratory responses. During DSE these estimations were used to calculate the vibratory responses. This method reduced the error between the surrogate and FEA from 85% to 17%. Lastly, a PCA-based multi-fidelity surrogate method was developed. This assumed the PCs of the high and low-fidelities were similar. The high-fidelity FE models had tens of thousands of nodes and the low-fidelity FE models had a few hundred nodes. The computational cost to create the surrogate was decreased by 75% for the same errors. For the same computational cost, the error was reduced by 50%. Together, the methods developed in this research were shown to decrease the cost of evaluating the structural responses of turbomachinery blade designs. They also provided a method to help the designer understand which designs to explore. This research paves the way for better, and more thoroughly understood turbomachinery blade designs.

Being able to accurately approximate outputs of computationally expensive simulations for arbitrary input parameters, also called missing points estimation, is central in many different areas of research and development with applications ranging from uncertainty propagation to control system design to name a few. This project investigates the potential of kernel transformations and nonlinear autoencoders as methods of improving the accuracy of the proper orthogonal decomposition method combined with regression. The techniques are applied on aerodynamic pressure CFD data around airplane wings in both two- and three-dimensional settings. The novel methods show potential in select situations, but cannot at this stage be generally considered superior. Their performances are similar although the procedure of design and training of a nonlinear autoencoder is less straight forward and more time demanding than using kernel transformations. The results demonstrate the regression bottleneck of the proper orthogonal decomposition method, which partially is improved with the new methods. Future studies should focus on adapting the autoencoder training strategy to the architecture and data as well as improving the regression stage of all methods.

The purpose of this project is to validate the adequacy of the Standard formula, used to calculate the Solvency Capital Requirement (SCR), with respect to a Swedish insurance company. The sub-modules evaluated are Equity risk (type 1) and Interest rate risk. The validation uses a quantitative assessment and the concept of Value at Risk (VaR). Additionally, investment strategies for risk free assets are evaluated through a scenario based analysis. The findings support that the Equity shock of 39%, as proposed in the Standard formula, is appropriate for a diversified portfolio of global equities. Furthermore, to some extent; the Equity shock is also sufficient for a diversified global portfolio with an overweight of Swedish equities. Additionally, the findings shows that the Standard formula for Interest rate risks occasionally underestimates the true Interest rate risk. Furthermore, it’s shown that there are some advantage of selecting an investment strategy that stabilizes the Own fund of an insurance company rather than a strategy that minimizes the SCR. / Syftet med detta arbete är att utvärdera Standardformeln, som används för att beräkna solvenskapitalkravet (SCR) under Solvens II, med avseende på dess lämplighet för ett svensk försäkringsbolag. Modulerna som utvärderas är aktierisk (typ 1) och ränterisk. Utvärderingen genomförs med kvantitativa metoder och utifrån konceptet Value at Risk (VaR). Dessutom utvärderas investeringsstrategier för riskfria tillgångar genom en scenariobaserad analys. Resultaten stödjer att den av Standardformeln föreskrivna aktiechocken på -39 % är tillräcklig för en diversifierad global aktieportfölj. Dessutom är aktiechocken även tillräcklig för en diversifierad global portfölj med en viss övervikt mot svenska aktier. Vidare visar resultaten att Standardformeln under vissa omständigheter underskattar ränterisken. Slutligen visar den scenariobaserade analysen att det är fördelaktigt att välja en investeringsstrategi som stabiliserar Own fund, hellre än en strategi som minimerar SCR.

Body odour perception plays an important role in human mate choice, especially in women. It was previously proposed that women select partners whose body odour resembles that of woman's fathers. Yet, this phenomenon has only been confirmed using ethological studies based on subjective perception of body odour similarities. Therefore, the aim of my diploma thesis was to test this hypothesis instrumentally using comprehensive GC×GC-TOFMS and subsequent multidimensional analyses of body odour chemical profiles of male partners and fathers of adult women. Body odour sampling from left and right axilla of fathers and partners of 41 women (altogether 164 samples) was performed using cotton swabs, which were then frozen and extracted into hexane. Typical human volatile substances, such as hydrocarbons, carboxylic acids, esters, alcohols, aldehydes, ketones, sterols and terpenes were detected in all examined samples. Using a newly available "tile-based" chromatographic alignment algorithm, we obtained a set of 341 compounds systematically occurring in male axillary odour. The principal component analysis was used to calculate Euclidean distances for all pairs of the studied male subjects. These estimates of "chemical distances" revealed to be significantly smaller for father-partner pairs of individual...

The idea behind this report is based on the fact that it is not only the number of users in the traffic network that is increasing, the number of connected devices such as probe vehicles and mobile sources has increased dramatically in the last decade. These connected devices provide large-scale mobility data and new opportunities to analyze the current traffic situation as they traverse through the network and continuously send out different types of information like Global Positioning System (GPS) data and Mobile Network Data (MND). Travel demand is often described in terms of an Origin Destination (OD) matrix which represents the number of trips from an origin zone to a destination zone in a geographic area. The aim of this master thesis is to develop and evaluate a data-driven method for estimation of dynamic OD matrices using unsupervised learning, sensor fusion and large-scale mobility data. Traditionally, OD matrices are estimated based on travel surveys and link counts. The problem is that these sources of information do not provide the quality required for online control of the traffic network. A method consisting of an offline process and an online process has therefore been developed. The offline process utilizes historical large-scale mobility data to improve an inaccurate prior OD matrix. The online process utilizes the results and tuning parameters from the offline estimation in combination with real-time observations to describe the current traffic situation. A simulation study on a toy network with synthetic data was used to evaluate the data-driven estimation method. Observations based on GPS data, MND and link counts were simulated via a traffic simulation tool. The results showed that the sensor fusion algorithms Kalman filter and Kalman filter smoothing can be used when estimating dynamic OD matrices. The results also showed that the quality of the data sources used for the estimation is of high importance. Aggregating large-scale mobility data as GPS data and MND by using the unsupervised learning method Principal Component Analysis (PCA) improves the quality of the large-scale mobility data and so the estimation results. / <p>Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet</p>

>Magister Scientiae - MSc / The Cape Floristic Region (CFR) in South Africa, is one of the world's five Mediterranean hotspots, and is also one of the 34 global biodiversity hotspots. It has rich biological diversity, high level of species endemism in flora and fauna and an unusual high level of human induced threats. The Sandveld forms part of the CFR and is also highly threatened by intensive agriculture (potato, rooibos and wheat farming), proliferation of tourism facilities, coastal development, and alien invasions. These biodiversity threats have led to habitat loss and are threatening the long-term security of surface and ground water resources. In order to understand trends in such biodiversity loss and improve in the management of these ecosystems, earth-orbiting observation satellite data were used. This research assessed landuse changes and trends in vegetation cover in the Sandveld, using remote sensing images. Landsat TM satellite images of 1990, 2004 and 2007 were classified using the maximum likelihood classifier into seven landuse classes, namely water, agriculture, fire patches, natural vegetation, wetlands, disturbed veld, and open sands. Change detection using remote sensing algorithms and landscape metrics was performed on these multi-temporal landuse maps using the Land Change ModelIer and Patch Analyst respectively. Markov stochastic modelling techniques were used to predict future scenarios in landuse change based on the classified images and their transitional probabilities. MODIS NDVI multi-temporal datasets with a 16day temporal resolution were used to assess seasonal and annual trends in vegetation cover using time series analysis (PCA and time profiling).Results indicated that natural vegetation decreased from 46% to 31% of the total landscape between 1990 and 2007 and these biodiversity losses were attributed to an increasing agriculture footprint. Predicted future scenario based on transitional probabilities revealed a continual loss in natural habitat and increase in the agricultural footprint. Time series analysis results (principal components and temporal profiles) suggested that the landscape has a high degree of overall dynamic change with pronounced inter and intra-annual changes and there was an overall increase in greenness associated with increase in agricultural activity. The study concluded that without future conservation interventions natural habitats would continue to disappear, a condition that will impact heavily on biodiversity and significant water dependent ecosystems such as wetlands. This has significant implications for the long-term provision of water from ground water reserves and for the overall sustainability of current agricultural practices.