In my thesis, I focus on the task of recognizing human iris from an image.In the beginning, the work deals with a question of biometrics, its importance and basic concepts, which are necessary for use in following text. Subsequently process of human Iris detection is described together with theory of evolution algorithms. In the implementation part, is described the design of implemented solution, which uses evolution algorithms, where is emphasis on correct pupil and iris boundary detection.

This thesis deals with one-document summarization of text data. Part of it is devoted to data preparation, mainly to the normalization. Listed are some of the stemming algorithms and it contains also description of lemmatization. The main part is devoted to Luhn"s method for summarization and its extension of use WordNet dictionary. Oswald summarization method is described and applied as well. Designed and implemented application performs automatic generation of abstracts using these methods. A set of experiments where developed, which verified correct functionality of the application and of extension of Luhn"s summarization method too.

Antonie Hofman's life and work The thesis will deal with by a remarkable personality of the franciscan terciary Antonie Hofman, who, as an example of his life and his literary work, still has something to say today. Her life is a testimony of personal bravery, modesty, but also of christian joy and the endless human strength and will resulting of solid faith in God. The content of the thesis will be the chronological breakdown of the individual phases of her life, which is crucial for understanding not only her life mission, but also her literary work. In addition, the work will list the publications she has written, with a brief evaluation. Attention is also paid to the catholic movement jocism, which was promoted by Antonie Hofman The work will also note Antonie Hofman's franciscan spirituality, which has shaped her life and work and which she has been able to master in the practice of everyday, often hard life.

There has been rapid change in the landcover/landuse in the Mekong delta, Viet Nam. The landcover/landuse has changed very fast due to intense population pressure, agriculture/aquaculture farming and timber collection in the coastal areas of the delta. The changing landuse pattern in the coastal areas of the delta is threatened to be flooded by sea level rise; sea level is expected to rise 33 cm until 2050; 45 cm until 2070 and 1 m until 2100. The coastline along the eastern Mekong delta has never been static, but the loss of mangrove forests along the coast has intensified coastline change. The objective of the present study is to map the changes in landcover/landuse along the eastern coast of the Mekong delta; and to detect the changes in position of the eastern coastline over the time period from 1989 to 2002.To detect changes in landuse, two satellite images of the same season, acquired by the TM sensor of Landsat 5 and the ETM+ sensor of Landsat 7 were used. The TM image was acquired on January 16, 1989 and ETM+ image was acquired on February 13, 2002. The landcover/landuse classes selected for the study are water, forest, open vegetation, soil and shrimp farms. Image differencing and post classification comparison are used to detect the changes between two time periods. Image to image correction technique is used to align satellite images. Maximum likelihood supervised classification technique is used to classify images. The result of the classification consists of five classes for 1989 and 2002, respectively. Overall accuracies of 87.5% and 86.8%, with kappa values of 0.85 and 0.84 are obtained for landuse 1989 and landuse 2002, respectively. The overall accuracy for the change map is 82% with kappa value 0.80. Post classification comparison is carried out in this study based on the supervised classification results. According to the results obtained from the post classification comparison, a significant decrease of 48% in forest and a significant increase of 74% in open vegetation and 21% in shrimp farms area observed over the entire study area. The coastline obtained by the combination of histogram thresholding and band ratio showed an overall advancement towards the South China Sea. The results showed that new land patches emerged along the eastern coast. The amount of new land patches appeared along the coast of the Mekong delta is approximately 2% of the entire study area.

Research in the area of public procurement has gained increasing attention with increasing interest in researching different evaluation methods for incoming tenders in procurement. The study analyzed the structure for procurement with a multi-criteria model for observation of changes in ranking and studied the importance of correct scoring. A sensitivity and robustness analysis were used as a tool to study the stability and the strength of a contracting entity's priorities. The purpose of this thesis was a study to determine if the ranking of tenders became more sensitive to the structure of the model or for each tender's evaluation, where it was found that the ability to correctly score points was more crucial than being critical of weights assigned to the criteria in the procurement, respectively. Relationship between price and quality turned out to be unknown to some extent, the connection could be determined after all tenders were submitted and accumulated points were available. / Forskning inom området offentlig upphandling har fått allt mer uppmärksamhet när intresset hamnat på undersökning av olika utvärderingsmetoder för inkommande anbud vid en upphandling. I studien analyserades uppbyggnaden för upphandlingar med en multikriteriemodell för observation över förändringar hos rangordning samt studerandet av vikten för en korrekt poängsättning. En känslighets- och robusthetsanalys tillämpades som verktyg för att studera stabiliteten och styrkan hos en upphandlande entitets prioriteringar. Syftet med uppsatsen var en undersökning för en bedömning om rangordningen av upphandlingars anbud blev känsligast mot strukturen på modellen eller för varje anbuds värdering, där det visade sig att förmågan för korrekt poängtilldelning var mer avgörande än att vara kritisk till vikterna som tilldelades respektive kriterier inom upphandlingen. Relationen mellan pris och kvalitet visade sig vara okänd till viss del, sambandet kunde avgöras efter att alla anbud var inskickade och samlade poäng fanns att tillhandahålla.

This document is trying to introduce the reader with issues of identifying the face connected with miscellaneous scanning technologies and enviroments. Inside this document there are mentioned some possibilities of creation unique print of a face so that there would be denied unwanted effects of enviroment and the identification of persons would be possible.

This diploma thesis focuses on the appropriate and effective use of film and serial adaptations of topics falling into the period of communist totalitarianism in Czechoslovakia (that is between 1948 and 1989, with emphasis on the period 1948-1969). The thesis is divided into four chapters, whereas the main part, which is didactic, is described in the third and fourth chapter. The first chapter describes the basic historical context of the communist government in Czechoslovakia between 1948 and 1989. It also shows and explains the origin of the idea of communism, the further development of the Communist Party of Czechoslovakia during the First Republic, and the postwar Third Republic. This chapter was put together using found edited sources and secondary literature. The second chapter deals with individual film and serial processing of topics related to the period. It presents specific films and series which were produced in the most recent years (or the post-revolutionary period), but also materials produced during the totalitarianism before 1989. In this chapter some works are discussed in greater detail, others are presented as selected alternative options for interaction for educational purposes. The third chapter summarizes the didactic and historical potential of the discussed topic, it also...

Les réseaux de neurones sont une famille de modèles de l'apprentissage automatique qui sont capable d'apprendre des tâches complexes directement des données. Bien que produisant déjà des résultats impressionnants dans beaucoup de domaines tels que la reconnaissance de la parole, la vision par ordinateur ou encore la traduction automatique, il y a encore de nombreux défis dans l'entraînement et dans le déploiement des réseaux de neurones. En particulier, entraîner des réseaux de neurones nécessite typiquement d'énormes ressources computationnelles, et les modèles entraînés sont souvent trop gros ou trop gourmands en ressources pour être déployés sur des appareils dont les ressources sont limitées, tels que les téléphones intelligents ou les puces de faible puissance. Les articles présentés dans cette thèse étudient des solutions à ces différents problèmes. Les deux premiers articles se concentrent sur l'amélioration de l'entraînement des réseaux de neurones récurrents (RNNs), un type de réseaux de neurones particulier conçu pour traiter des données séquentielles. Les RNNs sont notoirement difficiles à entraîner, donc nous proposons d'améliorer leur paramétrisation en y intégrant la normalisation par lots (BN), qui était jusqu'à lors uniquement appliquée aux réseaux non-récurrents. Dans le premier article, nous appliquons BN aux connections des entrées vers les couches cachées du RNN, ce qui réduit le décalage covariable entre les différentes couches; et dans le second article, nous montrons comment appliquer BN aux connections des entrées vers les couches cachées et aussi des couches cachée vers les couches cachée des réseau récurrents à mémoire court et long terme (LSTM), une architecture populaire de RNN, ce qui réduit également le décalage covariable entre les pas de temps. Nos expériences montrent que les paramétrisations proposées permettent d'entraîner plus rapidement et plus efficacement les RNNs, et ce sur différents bancs de tests. Dans le troisième article, nous proposons un nouvel optimiseur pour accélérer l'entraînement des réseaux de neurones. Les optimiseurs diagonaux traditionnels, tels que RMSProp, opèrent dans l'espace des paramètres, ce qui n'est pas optimal lorsque plusieurs paramètres sont mis à jour en même temps. A la place, nous proposons d'appliquer de tels optimiseurs dans une base dans laquelle l'approximation diagonale est susceptible d'être plus efficace. Nous tirons parti de l'approximation K-FAC pour construire efficacement cette base propre Kronecker-factorisée (KFE). Nos expériences montrent une amélioration en vitesse d'entraînement par rapport à K-FAC, et ce pour différentes architectures de réseaux de neurones profonds. Le dernier article se concentre sur la taille des réseaux de neurones, i.e. l'action d'enlever des paramètres du réseau, afin de réduire son empreinte mémoire et son coût computationnel. Les méthodes de taille typique se base sur une approximation de Taylor de premier ou de second ordre de la fonction de coût, afin d'identifier quels paramètres peuvent être supprimés. Nous proposons d'étudier l'impact des hypothèses qui se cachent derrière ces approximations. Aussi, nous comparons systématiquement les méthodes basées sur des approximations de premier et de second ordre avec la taille par magnitude (MP), et montrons comment elles fonctionnent à la fois avant, mais aussi après une phase de réapprentissage. Nos expériences montrent que mieux préserver la fonction de coût ne transfère pas forcément à des réseaux qui performent mieux après la phase de réapprentissage, ce qui suggère que considérer uniquement l'impact de la taille sur la fonction de coût ne semble pas être un objectif suffisant pour développer des bon critères de taille. / Neural networks are a family of Machine Learning models able to learn complex tasks directly from the data. Although already producing impressive results in many areas such as speech recognition, computer vision or machine translation, there are still a lot of challenges in both training and deployment of neural networks. In particular, training neural networks typically requires huge amounts of computational resources, and trained models are often too big or too computationally expensive to be deployed on resource-limited devices, such as smartphones or low-power chips. The articles presented in this thesis investigate solutions to these different issues. The first couple of articles focus on improving the training of Recurrent Neural Networks (RNNs), networks specially designed to process sequential data. RNNs are notoriously hard to train, so we propose to improve their parameterisation by upgrading them with Batch Normalisation (BN), a very effective parameterisation which was hitherto used only in feed-forward networks. In the first article, we apply BN to the input-to-hidden connections of the RNNs, thereby reducing internal covariate shift between layers. In the second article, we show how to apply it to both input-to-hidden and hidden-to-hidden connections of the Long Short-Term Memory (LSTM), a popular RNN architecture, thus also reducing internal covariate shift between time steps. Our experiments show that these proposed parameterisations allow for faster and better training of RNNs on several benchmarks. In the third article, we propose a new optimiser to accelerate the training of neural networks. Traditional diagonal optimisers, such as RMSProp, operate in parameters coordinates, which is not optimal when several parameters are updated at the same time. Instead, we propose to apply such optimisers in a basis in which the diagonal approximation is likely to be more effective. We leverage the same approximation used in Kronecker-factored Approximate Curvature (K-FAC) to efficiently build this Kronecker-factored Eigenbasis (KFE). Our experiments show improvements over K-FAC in training speed for several deep network architectures. The last article focuses on network pruning, the action of removing parameters from the network, in order to reduce its memory footprint and computational cost. Typical pruning methods rely on first or second order Taylor approximations of the loss landscape to identify which parameters can be discarded. We propose to study the impact of the assumptions behind such approximations. Moreover, we systematically compare methods based on first and second order approximations with Magnitude Pruning (MP), showing how they perform both before and after a fine-tuning phase. Our experiments show that better preserving the original network function does not necessarily transfer to better performing networks after fine-tuning, suggesting that only considering the impact of pruning on the loss might not be a sufficient objective to design good pruning criteria.

Marken är något fundamentalt. Jag dras till att skapa hålrum och att gjuta av dem. Låta händerna leda. Det är något som attraherar mig med att inte riktigt kunna föreställa mig formen av hålrummet i tanken, att det är ett glapp i överföringen mellan händernas rörelser och tankens. Att värdera det definierade högre än det odefinierade, vill jag lägga i en parantes.  Vi använder språk som materia, vi bygger betydelser, rum som skapar sammanhang, stödkonstruktioner, konstruktion som är dikt, som är en rörelse som är jord som är ett arbete som är sten som är en förflyttning som är trä som är en handling som är metall som är nära intill, att skulptera en tanke. Det är idén om att kroppen har en korrekt form som gör kroppen korrupt. Den brutna cirkeln är på ett sätt en helhet av två korrupta kroppar. Kategorier och definitioner är något verkligt, oavsett hur de är skapade eller kommit till. Vi uppmärksammar genast om något är konstigt. Två enheter som vi uppfattar som separata, är separata. Om de inte varit separata skulle vi uppfattat de som ett och inte som två. Det här ser vi omedelbart. Om vi inte är i ett främmande sammanhang med främmande definitioner, men då kommer säkerligen någon ganska snart att uppmärksamma oss på hur saker hänger ihop. Människor är hjälpsamma på sådant sätt. / My work may be described as bordering between sculpture and installation, intuitive and conceptual, it could in a way be seen as a broken circle, as one, but i will work hard to make the two entities, the two seemingly half circle shaped entities, to be separate, to be two.  I will assert they are one work but also not, I will assert their indepencence, they are two. The bridge, the connection, will (or will not) happen in the thought, not by sight. There will be no side-by-side, object dialog in near space, rather association through memory and through distance. Like an object or some kind of occurence is planted early on in a movie and then reoccurs later on with the purpuse to further the understanding of course of events.  We tend to read two half circles near each other as a whole circle. Robert Morris made two half circles with only a thin lit gap underlining their distinct form, but at the same time dissolving them as a broken circle. It is the idea of of a correct body that makes a body corrupt. The broken circle is in a way the whole of two corrupt bodies.

1. Einführung in die Theorie dynamischer Analysen mit Creo Simulate 2. Modalanalysen (Standard und mit Vorspannung) 3. Dynamische Analysen einschließlich Klassifizierung der Analysen; einige einfache Beispiele für eigene Studien (eine Welle unter Unwuchtanregung und ein Ein-Massen-Schwinger) sowie etliche Beispiele größerer dynamischer Systemmodelle aus unterschiedlichsten Anwendungsbereichen 4. Feedback an den Softwareentwickler PTC (Verbesserungsvorschläge und Softwarefehler) 5. Referenzen / 1. Introduction to dynamic analysis theory in Creo Simulate 2. Modal analysis (standard and with prestress) 3. Dynamic analysis, including analysis classification, some simple examples for own self-studies (shaft under unbalance excitation and a one-mass-oscillator) and several real-world examples of bigger dynamic systems 4. Feedback to the software developer PTC (enhancement requests and code issues) 5. References