Task-oriented dialogue systems are a popular way for organisations to generate extra value both internally and for customers. Modern approaches for these dialogue systems that use neural networks to enable training directly on written dialogues are very data hungry, which complicates their implementation. Crowdsourcing is an attractive solution for generating this type of training data, but the method also comes with several difficulties. We introduce a new method for generating training data based on parallel crowdsourcing of dialogues, as well as crowdsourced quality review. We use this method to collect a small dataset that takes place within the domain bus driver-traveler. We believe that this method offers an efficient way to collect new, high-quality datasets. Hybrid Code Networks is a model for dialogue systems that combines a neural network with domain-specific knowledge, and thus requires a significantly smaller amount of training data than other similar dialogue systems to achieve comparable performance. By combining Hybrid Code Networks with our new method for generating training data, we believe that the threshold for implementing task-oriented dialogue systems on domains with insufficient training data can be lowered. We implement Hybrid Code Networks and train the implementation on the collected dataset and achieve good results. / Uppgiftsorienterade dialogsystem är ett populärt sätt för företag att generera extra värde både internt och för kunder. Moderna modeller för dessa dialogsystem som använder neurala nätverk för att möjliggöra träning direkt på skriftliga dialoger är väldigt datahungriga, vilket försvårar implementationen av dessa. Crowdsourcing är en attraktiv lösning för att generera denna typ av träningsdata, men metoden kommer även med flera svårigheter. Vi introducerar en ny metod för generering av träningsdata som bygger på parallell crowdsourcing av dialoger, samt crowdsourcad kvalitetsgranskning. Vi använder denna metod för att samla in ett litet dataset som utspelar sig inom domänen busschaufför-resenär. Vi menar att denna metod erbjuder ett effektivt sätt att samla in nya, högkvalitativa dataset. Hybrid Code Networks är en modell för dialogsystem som kombinerar ett neuralt nätverk med domänspecifik kunskap, och som på så sätt kräver en betydligt mindre mängd träningsdata än andra liknande dialogsystem för att uppnå jämförbar prestanda. Genom att kombinera Hybrid Code Networks med vår nya metod för generering av träningsdata menar vi att man kan sänka tröskeln för att implementera uppgiftsorienterade dialogsystem på domäner med otillräcklig träningsdata. Vi implementerar Hybrid Code Networks och tränar implementationen på det insamlade datasetet, och uppnår goda resultat.

This work focuses on the analysis of plasma irregularities generated during two active space experiments: the injection of an artificial dust layer, and high-power radio waves. The objective of the "first experiment is to examine the effects of artificially created dust layers on the scatter of radars from plasma irregularities embedded in dusty plasma in space. This is an alternate approach for understanding the mechanisms of enhanced radar scatter from plasma irregularities embedded in Noctilucent Clouds and Polar Mesospheric Summer Echoes. The second experiment involves a transmission of high power electromagnetic waves into the ionospheric plasma from the ground, which can excite stimulated electromagnetic emissions offset from the transmitter frequency. These stimulated electromagnetic emissions provide diagnostic information of the ionosphere and thus can be used to investigate fundamental physical principles which govern the earth\'s ionosphere, so that present and future transmission technologies may take into account the complexities of the ionosphere. The interaction altitude of the artificial dust layer and high power radio waves is approximately 250 km and 160 km respectively, thus dealing with uniquely different regions of the ionosphere. Each experiment is discussed separately using theoretical, observational and advanced computational methodologies. The study first investigates plasma turbulence associated with the creation of an artificial dust layer in the earth's ionosphere. Two scenarios are considered for plasma irregularity generation as dust is injected at an oblique angle across the geomagnetic field. The first is a shear-driven plasma instability due to inhomogeneities in the boundary layer between the injected charged dust layer and the background plasma. This begins to appear at very early times once the dust is released into the space plasma, which is of the order or less than the dust charging time period. The second mechanism is free streaming of the charged dust relative to the background plasma. This produces irregularities at times much longer than the dust charging period and also longer than the dust plasma period. Although both mechanisms are shown to produce turbulence in the lower hybrid frequency range, the resulting irregularities have important differences in their physical characteristics. A comparison between the processes is made to determine the consequences for upcoming observations. Both processes are shown to have the possibility of generating turbulence after the release of dust for the regimes of upcoming space experiments over a range of timescales. This work also presents the first observations of unique narrowband emissions ordered near the Hydrogen ion (H+) gyro-frequency (fcH) in the Stimulated Electromagnetic Emission (SEE) spectrum when the transmitter is tuned near the second electron gyro-harmonic frequency (2fce), during ionospheric modification experiments. The frequency structuring of these newly discovered emission lines is quite unexpected since H+ is known to be a minor constituent in the interaction region which is near 160 km altitude. The spectral lines are typically shifted from the pump wave frequency by harmonics of a frequency about 10% less than fcH (" 800 Hz) and have a bandwidth of less than 50 Hz which is near the O+ gyro-frequency fcO. A theory is proposed to explain these emissions in terms of a Parametric Decay Instability (PDI) in a multi-ion species plasma due to possible proton precipitation associated with the disturbed conditions during the heating experiment. The observations can be explained by including several percent H+ ions into the background plasma. The implications are new possibilities for characterizing proton precipitation events during ionospheric heating experiments. / Ph. D.

Ce manuscrit présente trois expériences menées dans le cadre de la fusion nucléaire par confinement inertiel. La première expérience porte sur l'étude la propagation d'un faisceau d'électrons rapides dans un plasma pré-comprimé dans le cadre du schémas d'allumage rapide. Deux expériences sur la génération d'onde de choc dans des plasmas ont été menées dans le cadre du schéma d'allumage par choc. La première expérience a été consacre à l'étude le la propagation d'un faisceau d'électrons rapides dans une cible comprimée. L'implosion de la cible avec une géométrie cylindrique a été menée avec l'installation laser GEKKO XII (ILE Osaka, Japon). Le faisceau d'électrons rapides a été injecté en utilisant l'installation laser LFEX (_1019W/cm2). et sa propagation à travers le cylindre comprimé a été observée avec plusieurs diagnostics X. Cette expérience a démontré l'effet d'un guidage du faisceau d'électrons rapides résultant du champ magnétique auto-généré. Par ailleurs, les résultats de cette expérience ont été en bon accord avec deux des simulation numériques. Cette étude a fait l'objet de la publication Approach to the study of fast electron transport in cylindrically imploded targets, Laser and particle Beams, 33, 525-534, (2015). Deux autres expériences ont été réalisées pour l’étude de la propagation de chocs forts créés par laser dans un plasma. Celles-ci ont été réalisées avec différentes installations laser.Dans la première expérience avec le laser Gekko Xll, nous avons observé la création et la propagation de deux ondes de choc successives dans des plasmas d’ablation de CH et de Be. L’objectif de caractériser l’amplification d’un choc transmis par la collision des deux chocscontre propagatifs a été partiellement réalisé. La comparaison des résultats expérimentauxà des simulations hydrodynamiques a permis d’établir une amplification du choc d’unfacteur 2 en pression pour les conditions expérimentales réalisées. Les tirs sur une cible deBe a permis de développer et valider les outils d’exploitation de la propagation de deuxchocs par radiographie X. La deuxième expérience a été réalisée avec laser PHELIX de GSI(Darmstadt, Allemagne). Le but de cette expérience était d’étudier la génération de chocsforts qui ont été utilisés pour étudier l’équations d’état du carbone dans le domaine WDMpertinent pour le planétologie. Les conditions de pression et de densité pour le carboneont été obtenues en déduisant la pression et la vitesse de l’onde de choc des diagnosticschronométriques employés dans cette expérience. Des états du diamant en phase liquidemétallique pour des pressions de l’ordre de 7 Mbar et des températures de 15,000 degrésont été obtenus. / This manuscript presents three experiments conducted as part of a nuclear fusion byinertial confinement. The first experiment is the study of the fast electron beam propagationin a pre-compressed plasma in the fast ignition scheme. Two other experiments about theshock wave generation in plasmas were conducted in the ignition shock pattern.The first experiment was devoted to the study of the fast electron beam transport in a compressed target. The implosion of the target with a cylindrical geometry was carriedout with the GEKKO XII laser facility (ILE Osaka, Japan). The fast electron beam wasgenerated by the LFEX laser ( 1019W/cm2) and its propagation through the compressedcylinder was observed with several X-ray diagnostics. This experiment showed the guidingeffect of the electron beam resulting from self-generated magnetic fields. Furthermore, theresults of this experiment were in good agreement with numerical simulations. This studywas the subject of the publication Approach to the study of fast electron transport incylindrically imploded targets, Laser and Particle Beams, 33,525-534,(2015). Two other experiments were performed to study the propagation of strong shockscreated by lasers in a plasma. These were carried out with different laser systems. In the firstexperiment with the Gekko XII laser, we observed the creation and the propagation of twosuccessive shock waves in an ablation plasma in CH and Be. The objective to characterizethe amplification of a transmitted shock by the collision of two counter propagatingshocks has been partially realized. The comparison of the experimental results with thehydrodynamic simulations enabled us to confirm an amplification of the shock by a factor2 in pressure in the condition of this experiment. The shot with a Be target allowed todevelop and to validate the diagnostic method of X-ray radiography for shock propagation.The second experiment was performed with laser PHELIX GSI (Darmstadt, Germany).The purpose of this experiment was to study the generation of strong shocks. They wereapplied to study the equation of state of carbone in WDM state for the planetology. Thecondition of pressure and density for the carbon were obtained by deducting the pressureand the velocity of the shock wave chronometric diagnostics employed in this experiment.In this experiment, diamond was at the metallic liquid phase with the pressure of 7 Mbarand the temperature of 15,000 degrees

<p>This dissertation considers packet-switched wireless networks for transmission of variable-rate layered hybrid video streams. Target applications are video streaming and broadcasting services. The work can be divided into two main parts.</p><p>In the first part, a novel quality-scalable scheme based on coefficient refinement and encoder quality constraints is developed as a possible extension to the video coding standard H.264. After a technical introduction to the coding tools of H.264 with the main focus on error resilience features, various quality scalability schemes in previous research are reviewed. Based on this discussion, an encoder decoder framework is designed for an arbitrary number of quality layers, hereby also enabling region-of-interest coding. After that, the performance of the new system is exhaustively tested, showing that the bit rate increase typically encountered with scalable hybrid coding schemes is, for certain coding parameters, only small to moderate. The double- and triple-layer constellations of the framework are shown to perform superior to other systems.</p><p>The second part considers layered code streams as generated by the scheme of the first part. Various error propagation issues in hybrid streams are discussed, which leads to the definition of a decoder quality constraint and a segmentation of the code stream to transmit. A packetization scheme based on successive source rate consumption is drafted, followed by the formulation of the channel code rate optimization problem for an optimum assignment of available codes to the channel packets. Proper MSE-based error metrics are derived, incorporating the properties of the source signal, a terminate-on-error decoding strategy, error concealment, inter-packet dependencies, and the channel conditions. The Viterbi algorithm is presented as a low-complexity solution to the optimization problem, showing a great adaptivity of the joint source channel coding scheme to the channel conditions. An almost constant image qualiity is achieved, also in mismatch situations, while the overall channel code rate decreases only as little as necessary as the channel quality deteriorates. It is further shown that the variance of code distributions is only small, and that the codes are assigned irregularly to all channel packets.</p><p>A double-layer constellation of the framework clearly outperforms other schemes with a substantial margin. </p><p>Keywords — Digital lossy video compression, visual communication, variable bit rate (VBR), SNR scalability, layered image processing, quality layer, hybrid code stream, predictive coding, progressive bit stream, joint source channel coding, fidelity constraint, channel error robustness, resilience, concealment, packet-switched, mobile and wireless ATM, noisy transmission, packet loss, binary symmetric channel, streaming, broadcasting, satellite and radio links, H.264, MPEG-4 AVC, Viterbi, trellis, unequal error protection</p>

This dissertation considers packet-switched wireless networks for transmission of variable-rate layered hybrid video streams. Target applications are video streaming and broadcasting services. The work can be divided into two main parts. In the first part, a novel quality-scalable scheme based on coefficient refinement and encoder quality constraints is developed as a possible extension to the video coding standard H.264. After a technical introduction to the coding tools of H.264 with the main focus on error resilience features, various quality scalability schemes in previous research are reviewed. Based on this discussion, an encoder decoder framework is designed for an arbitrary number of quality layers, hereby also enabling region-of-interest coding. After that, the performance of the new system is exhaustively tested, showing that the bit rate increase typically encountered with scalable hybrid coding schemes is, for certain coding parameters, only small to moderate. The double- and triple-layer constellations of the framework are shown to perform superior to other systems. The second part considers layered code streams as generated by the scheme of the first part. Various error propagation issues in hybrid streams are discussed, which leads to the definition of a decoder quality constraint and a segmentation of the code stream to transmit. A packetization scheme based on successive source rate consumption is drafted, followed by the formulation of the channel code rate optimization problem for an optimum assignment of available codes to the channel packets. Proper MSE-based error metrics are derived, incorporating the properties of the source signal, a terminate-on-error decoding strategy, error concealment, inter-packet dependencies, and the channel conditions. The Viterbi algorithm is presented as a low-complexity solution to the optimization problem, showing a great adaptivity of the joint source channel coding scheme to the channel conditions. An almost constant image qualiity is achieved, also in mismatch situations, while the overall channel code rate decreases only as little as necessary as the channel quality deteriorates. It is further shown that the variance of code distributions is only small, and that the codes are assigned irregularly to all channel packets. A double-layer constellation of the framework clearly outperforms other schemes with a substantial margin. Keywords — Digital lossy video compression, visual communication, variable bit rate (VBR), SNR scalability, layered image processing, quality layer, hybrid code stream, predictive coding, progressive bit stream, joint source channel coding, fidelity constraint, channel error robustness, resilience, concealment, packet-switched, mobile and wireless ATM, noisy transmission, packet loss, binary symmetric channel, streaming, broadcasting, satellite and radio links, H.264, MPEG-4 AVC, Viterbi, trellis, unequal error protection

Cette thèse s'inscrit dans le cadre de la recherche sur la fusion nucléaire par confinement inertiel, et vise notamment à contribuer à la validation du schéma d'allumage rapide. Elle consiste en une étude expérimentale des différents processus impliqués dans la propagation d'un faisceau d'électrons relativistes, produit par une impulsion laser ultra-intense (10^{19} W.cm-2), au sein de la matière dense qu'elle soit solide ou comprimée. Dans ce travail de recherche nous présentons les résultats de trois expériences réalisées sur des installations laser distinctes afin de générer des faisceaux d'électrons dans diverses conditions et d'étudier leur propagation dans différents états de la matière, du solide froid au plasma dense et tiède.La première expérience a été réalisée à très haut contraste temporel sur l'installation laser UHI100 du CEA de Saclay. L'étude du dépôt d'énergie par le faisceau d'électrons dans l'aluminium solide a mis en évidence un important chauffage à faible profondeur, où les effets collectifs sont prédominants, générant ainsi un gradient important de température entre les faces avant (300eV) et arrière (20eV) sur 20µm d'épaisseur. Une modélisation numérique de l'expérience montre que ce gradient induit la formation d'une onde de choc débouchant en face arrière, donnant alors lieu à une augmentation de l'émission thermique. La chronométrie expérimentale du débouché du choc permet de valider le modèle de transport collectif des électrons.Deux autres expériences ont porté sur l'étude de la propagation de faisceaux d'électrons rapides au sein de cibles comprimées. Lors de la première expérience sur LULI2000 (LULI), la géométrie de compression plane a permis de dissocier de manière précise les pertes d'énergie liées aux effets résistifs de celles liées aux effets collisionnels. En comparant nos résultats expérimentaux à des simulations, nous avons mis en évidence l'augmentation significative des pertes d'énergie du faisceau d'électrons avec la compression et le chauffage de la cible à des température proches de la température de Fermi, et ce, pour les deux mécanismes. La seconde expérience, réalisée en géométrie cylindrique sur Vulcan (RAL), a permis de mettre en évidence un phénomène de guidage du faisceau d'électrons rapides sous l'effet d'un intense champ magnétique, auto-généré en présence d'importants gradients radiaux de résistivité. Par ailleurs, dans les conditions de température et de densité atteintes, l'augmentation des pertes d'énergie collisionnelles avec la densité s'avère être compensée par une diminution des pertes résistives du fait du passage de la conductivité du milieu dans le régime des hautes températures de Spitzer. / The framework of this PhD thesis is the validation of the fast ignition scheme for the nuclear fusion by inertial confinement. It consists in the experimental study of the various processes involved in fast electron beams propagation, produced by intense laser pulses (10^{19} W.cm-2), through dense matter either solid or compressed. In this work we present the results of three experiments carried out on different laser facilities in order to generate fast electron beams in various conditions and study their propagation in different states of matter, from the cold solid to the warm and dense plasma.The first experiment was performed with a high intensity contrast on the UHI100 laser facility (CEA Saclay). The study of fast electron energy deposition inside thin aluminium targets highlights a strong target heating at shallow depths, where the collectivs effects are predominant, thus producing a steep temperature profile between front (300eV) and rear (20eV) sides over 20µm thickness. A numerical simulation of the experiment shows that this temperature gradient induces the formation of a shock wave, breaking through the rear side of the target and thus leading to increase the thermal emission. The experimental chronometry of the shock breakthrough allowed validating the model of the collective transport of electrons.Two other experiments were dedicated to the study of fast electron beam propagation inside compressed targets. In the first experiment on the LULI2000 laser facility, the plane compression geometry allowed to precisely dissociate the energy losses due to resistive effects from those due to the collisional ones. By comparing our experimental results with simulations, we observed a significative increase of the fast electron beam energy losses with the compression and the target heating to temperatures close to the Fermi temperature. The second experiment, performed in a cylindrical geometry, demonstrated a fast electron beam guiding phenomenon due to self-generated magnetic fields in presence of sharp radial resistivity gradients. Furthermore, in the temperature and density conditions achieved here, the increase of collisional energy losses with density is compensated by the decreasing resistive energy losses due to the transition of the conductivity into the high-temperatures Spitzer regime.