No description available.

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.

The presented diploma thesis was elaborated as an architectural study of a Central European Forum in Olomouc (SEFO). Campus SEFO will be created as an reconstruction of the Museum of Modern Art (MUO) in Denis street and building in a neighboring vacant lot. The proposal involves urban, architectural, operational layout, design and material solutions objects in spatial context. Within SEFO and MUO they are created each operation - stand-alone units. Objects SEFO and MUO are interconnected. It is necessary to respect the separation of publicly accessible areas of compartments accessible only by employees. Architectural study includes space for exhibitions, library, multi-purpose space with facilities, vestibule usable for exhibition openings and other cultural activities, chamber music performances, as well as facilities for education, technological facilities of the building, the depositary (transport and central), photo studio restoration studio, office space, locker rooms and restrooms personnel. SEFO specific aim of capturing the diverse manifestations of visual culture of Central Europe after World War 2, the building's permanent exhibition, acquisition activity, temporary exhibitions, including larger medium-shows (eg. The biennial or triennial), discussion forums and other supporting cultural events.

This Master Thesis is the research done in the country of Rwanda. The project leads to study the climate of this country in order to establish whether this climate could be used to produce energy from air and to implement the first wind turbine for serving the nation.   After an introduction about the historical background of wind power, the thesis work deals with assessment of wind energy potential of Rwanda in focusing of the most suitable place for wind power plants. The best location with annual mean wind speed, the rate of use of turbine with hub height for an annual production per year, the mean wind speeds for 6 sites of Rwanda based on ECMWF for climatic data for one year at relief of altitude of 100m and coordinates are reported too.   The result of energy produced and calculations were done based on power hitting wind turbine generator in order to calculate Kinetic energy and power available at the best location to the measurement over the period of 12 months, that could be hoped for long term.   With help of logarithmic law, where wind speed usually increases with increasing in elevation and the desired wind speeds at all 6 sites were used. The annual energy production was taken into account at the best site with desired wind speed at the initial cost of turbine as well as the cost of energy (COE).However, with comparison of the tariff of EWSA, the price of Wind designed in this Research per kWh is cheaper and suitable for people of Rwanda. / <p><em>Rwanda has considerable opportunities development energy from hydro sources, methane gas, solar and peat deposits. Most of these energy sources have not been fully exploited, such as solar, wind and geothermal. As such wood is still being the major source of energy for 94 per cent of the population and imported petroleum products consume more than 40 per cent of foreign exchange. Energy is a key component of the Rwandan economy. It is thus recognized that the current inadequate and expensive energy supply constitutes a limiting factor to sustainable development. Rwanda’s Vision 2020 emphasizes the need for economic growth, private investment and economic transformation supported by a reliable and affordable energy supply as a key factor for the development process. To achieve this transformation, the country will need to increase energy production and diversify into alternative energy sources. Rwandan nations don’t have small-scale solar, wind, and geothermal devices in operation providing energy to urban and rural areas. These types of energy production are especially useful in remote locations because of the excessive cost of transporting electricity from large-scale power plants. The application of renewable energy technology has the potential to alleviate many of the problems that face the people of Rwanda every day, especially if done so in a sustainable manner that prioritizes human rights.</em></p>