Дисертації з теми "Coherent drive"

This dissertation considers both a lossless and a dissipative micromaser in which a monoenergetic beam of two-level atoms in a coherent superposition of their upper and lower states is injected inside a single mode, high-Q cavity. In the lossless case, we find that under appropriate conditions a field initially in a mixed state will evolve to previously unknown pure states, which we call the tangent and cotangent states. In various limits, these states exhibit interesting properties such as sub-Poissonian photon statistics and squeezing, and most importantly they acquire the characteristics of "macroscopic" quantum superpositions. When dissipation of the cavity mode is incorporated into the model, we find that although the field no longer evolves to a pure state, the mixed steady-state field may still retain the properties of a macroscopic superposition under experimentally realizable damping rates. We then evaluate the experimental conditions necessary for the preparation and detection of such macroscopic superpositions.

In the present dissertation paper we study problems related to synchronization phenomena in the presence of noise which unavoidably appears in real systems. One part of the work is aimed at investigation of utilizing delayed feedback to control properties of diverse chaotic dynamic and stochastic systems, with emphasis on the ones determining predisposition to synchronization. Other part deals with a constructive role of noise, i.e. its ability to synchronize identical self-sustained oscillators. First, we demonstrate that the coherence of a noisy or chaotic self-sustained oscillator can be efficiently controlled by the delayed feedback. We develop the analytical theory of this effect, considering noisy systems in the Gaussian approximation. Possible applications of the effect for the synchronization control are also discussed. Second, we consider synchrony of limit cycle systems (in other words, self-sustained oscillators) driven by identical noise. For weak noise and smooth systems we proof the purely synchronizing effect of noise. For slightly different oscillators and/or slightly nonidentical driving, synchrony becomes imperfect, and this subject is also studied. Then, with numerics we show moderate noise to be able to lead to desynchronization of some systems under certain circumstances. For neurons the last effect means “antireliability” (the “reliability” property of neurons is treated to be important from the viewpoint of information transmission functions), and we extend our investigation to neural oscillators which are not always limit cycle ones. Third, we develop a weakly nonlinear theory of the Kuramoto transition (a transition to collective synchrony) in an ensemble of globally coupled oscillators in presence of additional time-delayed coupling terms. We show that a linear delayed feedback not only controls the transition point, but effectively changes the nonlinear terms near the transition. A purely nonlinear delayed coupling does not affect the transition point, but can reduce or enhance the amplitude of collective oscillations.
In dieser Dissertation werden Synchronisationsphänomene im Vorhandensein von Rauschen studiert. Ein Ziel dieser Arbeit besteht in der Untersuchung der Anwendbarkeit verzögerter Rückkopplung zur Kontrolle von bestimmten Eigenschaften chaotischer oder stochastischer Systeme. Der andere Teil beschäftigt sich mit den konstruktiven Eigenschaften von Rauschen. Insbesondere wird die Möglichkeit, identische selbsterregte Oszillatoren zu synchronisieren untersucht. Als erstes wird gezeigt, dass Kohärenz verrauschter oder chaotischer Oszillatoren durch verzögertes Rückkoppeln kontrolliert werden kann. Es wird eine analytische Beschreibung dieses Phänomens in verrauschten Systemen entwickelt. Außerdem werden mögliche Anwendungen im Zusammenhang mit Synchronisationskontrolle vorgestellt und diskutiert. Als zweites werden Oszillatoren unter dem Einfluss von identischem Rauschen betrachtet. Für schwaches Rauschen und genügend glatte Systeme wird bewiesen, das Rauschen zu Synchronisation führt. Für leicht unterschiedliche Oszillatoren und leicht unterschiedliches Rauschen wird die Synchronisation unvollständig. Dieser Effekt wird auch untersucht. Dann wird mit Hilfe von Numerik gezeigt, dass moderates Rauschen zur Desynchronisierung von bestimmten Systemen führen kann. Dieser Effekt wird auch in neuronalen Oszillatoren untersucht, welche nicht unbedingt Grenzzyklen besitzen müssen. Im dritten Teil wird eine schwache nichtlineare Theorie des Kuramoto-Übergangs, dem Übergang zur kollektiven Synchronisation, in einem Ensemble von global gekoppelten Oszillatoren mit zusätzlichen zeitverzögerten Kopplungstermen entwickelt. Es wird gezeigt, dass lineare Rückkopplung nicht nur den Übergangspunkt bestimmt, sondern auch die nichtlinearen Terme in der Nähe des Übergangs entscheidend verändert. Eine rein nichtlineare Rückkopplung verändert den Übergang nicht, kann aber die Amplitude der kollektiven Oszillationen vergrößern oder verringern.

We consider a pair of uncoupled conditional oscillators near a subcritical Hopf bifurcation that are driven by two weak white noise sources, one intrinsic and one common. The effect of the competition between the common and intrinsic noise forcing on the synchronization behaviour of the phases of these two oscillators is studied. Using a stochastic multiple scales method, we derive the envelope equations of the oscillators and then use the theory of linearized stochastic differential equations as well as an asymptotic analysis to study the probability density of the phase difference of the oscillators. It is found that common noise increases the degree of synchrony in the pair of oscillators and that it can be characterized by the ratio of intrinsic to common noise. Furthermore, the nonlinear dynamics of the oscillators can affect the character of this synchronization in terms of the average phase difference. The results are also related to the study of spike time reliability and possible implications are briefly discussed.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2013.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 189-195).
Coherent hard x-rays have many medical, commercial and academic research applications. To facilitate the design of a table-top coherent hard x-ray source, this thesis studies the linear acceleration of electrons by optical lasers in unbounded vacuum, the linear acceleration and compression of electron bunches by coherent terahertz pulses in cylindrical waveguides, and the generation of coherent hard x-ray radiation by nonlinear Thomson scattering of compressed electron bunches. The Lawson-Woodward theorem describes conditions prohibiting net electron acceleration in laser-electron interactions. We point out how the Lawson-Woodward theorem permits net linear acceleration of a relativistic electron in unbounded vacuum and verify this with electrodynamic simulations. By hypothesizing that substantial net linear acceleration is contingent on the field's ability to bring the particle to a relativistic energy in its initial rest frame, we derive a general formula for the acceleration threshold, which is useful as a practical guide to the laser intensities that linear vacuum acceleration requires. We characterize the scaling laws of linear acceleration by a pulsed radially-polarized beam in infinite vacuum, showing that greater energy gain is achievable with tighter focusing and the use of pre-accelerated electrons. We propose a two-color linear acceleration scheme that exploits changes in the interference pattern caused by the Gouy phase shift to achieve over 90% the one-color theoretical gain limit, more than twice the 40% achievable with a one-color paraxial beam. Interested in capitalizing on the larger wavelengths of coherent terahertz radiation to accelerate larger electron bunches, we study electron acceleration and bunch compression in a cylindrical metal-coated dielectric waveguide. We numerically predict an achievable acceleration gradient of about 450 MeV/m using a 20 mJ terahertz pulse, and separately achieve a 50 times compression to a few-femtosecond duration of a 1.6 pC relativistic electron bunch. Finally, we numerically study the production of coherent hard x-rays via nonlinear Thomson scattering for different degrees of laser focusing. We derive an approximate analytical formula for the optimal incident field intensity that maximizes the radiation intensity spectral peak for a given output and input frequency.
by Liang Jie Wong.
Ph.D.

Thesis (Ph. D.)--University of California, San Diego, 2008.
Title from first page of PDF file (viewed Sept. 9, 2008). Available via ProQuest Digital Dissertations. Vita. Includes bibliographical references (p. 202-209).

Summarization models which operate on meaning representations of documents have been neglected in the past, although they are a very promising and interesting class of methods for summarization and text understanding. In this thesis, I present one such summarizer, which uses the proposition as its meaning representation. My summarizer is an implementation of Kintsch and van Dijk's model of comprehension, which uses a tree of propositions to represent the working memory. The input document is processed incrementally in iterations. In each iteration, new propositions are connected to the tree under the principle of local coherence, and then a forgetting mechanism is applied so that only a few important propositions are retained in the tree for the next iteration. A summary can be generated using the propositions which are frequently retained. Originally, this model was only played through by hand by its inventors using human-created propositions. In this work, I turned it into a fully automatic model using current NLP technologies. First, I create propositions by obtaining and then transforming a syntactic parse. Second, I have devised algorithms to numerically evaluate alternative ways of adding a new proposition, as well as to predict necessary changes in the tree. Third, I compared different methods of modelling local coherence, including coreference resolution, distributional similarity, and lexical chains. In the first group of experiments, my summarizer realizes summary propositions by sentence extraction. These experiments show that my summarizer outperforms several state-of-the-art summarizers. The second group of experiments concerns abstractive generation from propositions, which is a collaborative project. I have investigated the option of compressing extracted sentences, but generation from propositions has been shown to provide better information packaging.

Le traitement optique de l'information quantique nécessite des émetteurs de photons uniques indiscernables. Dans ce cadre, des émetteurs quantiques récemment découverts dans les matériaux 2D offrent de nouvelles perspectives en termes de dispositifs photoniques intégrés. Ainsi, dans le nitrure de bore hexagonal (hBN), une nouvelle famille de centres colorés a l'avantage de posséder une faible dispersion en longueur d'onde. Ces centres émettant dans le bleu (λ ≈ 435 nm) peuvent également être positionnés de manière déterministe. Ces deux qualités sont rares parmi les émetteurs quantiques dans l'état solide et s'ajoutent en outre à des propriétés photophysiques avantageuses. Cette famille d'émetteurs constitue l'objet d'étude principal de cette thèse.Dans un premier temps, nous détaillons les principales figures de mérite d'un émetteur de photons uniques : brillance, pureté, cohérence temporelle et indiscernabilité. Nous évoquerons également les principaux systèmes physiques émettant des photons uniques, afin de contextualiser la caractérisation à suivre des centres colorés bleus dans hBN.Nous décrivons, dans une deuxième partie, les méthodes expérimentales génériques employées au cours de la thèse en commençant par l'exfoliation mécanique permettant d'obtenir des cristaux de hBN et l'irradiation électronique pour la création des centres colorés. Ceux-ci sont ensuite caractérisés optiquement à l'échelle individuelle au moyen de techniques combinant microscopie confocale, cryogénie, comptage de photons et spectroscopie. Nous détaillons également le traitement des données utilisé pour calculer la fonction d'autocorrélation d'intensité. Le troisième chapitre est consacré aux mesures de différentes propriétés photophysiques des centres bleus à l'échelle de l'émetteur individuel, telles que le temps de vie, la pureté, la polarisation et la photostabilité. Nous nous intéressons également au processus de création des centres colorés bleus, en effectuant des mesures de cathodoluminescence in situ, complétées par des mesures optiques. La nature microscopique de cette famille de centres colorés est également évoquée.Nous traitons ensuite de l'excitation laser résonante d'un centre bleu. L'étude des corrélations de photons permet d'observer des oscillations de Rabi, et d'en extraire le temps de cohérence de l'émetteur. En outre, ces corrélations donnent accès à la dynamique de la diffusion spectrale prenant place à une échelle de temps de l'ordre de la dizaine de microsecondes. Enfin, nous étudions l'indiscernabilité des photons émis par un centre bleu en mesurant les corrélations de photons dans un interféromètre de Hong, Ou et Mandel. Nous mettons en évidence le phénomène d'interférence à deux photons témoignant de l'indiscernabilité partielle des photons émis par le centre coloré. Ce résultat prometteur pourra être améliorée grâce à l'intégration des émetteurs dans des structures photoniques visant à augmenter la collection et diminuer l'impact du déphasage.Les résultats détaillés dans cette thèse démontrent le potentiel de ces centres colorés bleus dans hBN pour des applications dans le domaine de l'information quantique. De futur développements permettront une meilleure compréhension et un meilleur contrôle de leur dynamique d'émission ainsi que leur intégration dans des dispositifs optoélectroniques. Ces travaux ouvrent de nouvelles perspectives en termes de photonique quantique avec des matériaux 2D
Optical quantum information processing requires single and indistinguishable photon emitters. In this context, recently discovered quantum emitters in 2D materials offer new perspectives in terms of integrated photonic devices. In hexagonal boron nitride (hBN), a new family of color centers has the advantage of a low wavelength dispersion. These blue-emitting color centers (λ ≈ 435 nm) can also be positioned deterministically. These two qualities are rare among solid-state quantum emitters, and add up to advantageous photophysical properties. This family of emitters is the main focus of this thesis.First, we detail the main figures of merit of a single-photon emitter: brightness, purity, temporal coherence and indistinguishability. We also discuss the main physical systems emitting single photons, in order to contextualize the following characterization of blue color centers in hBN.In the second part, we describe the generic experimental methods used during the thesis: mechanical exfoliation to obtain hBN crystals and electron irradiation to create the color centers. The latter are then optically characterized on an individual scale using techniques combining confocal microscopy, cryogenics, photon counting and spectroscopy. Finally, we describe the data processing methods used to calculate the intensity autocorrelation function.The third chapter is devoted to measurements of various photophysical properties of blue centers at the scale of individual emitters, such as lifetime, purity, polarization and photostability. We also focus on the creation process of blue color centers, by carrying out in-situ cathodoluminescence measurements, complemented by optical measurements. The microscopic nature of this family of color centers is also discussed.We then address the resonant laser excitation of a blue center. The study of photon correlations allows us to observe Rabi oscillations, and to extract the coherence time of the emitter. In addition, these correlations give access to the dynamics of the spectral diffusion taking place on a time scale of a few tenths of microseconds. Finally, we study the indistinguishability of photons emitted by a blue center by measuring photon correlations in a Hong-Ou-Mandel interferometer. We demonstrate two-photon interference, indicating partial indistinguishability of photons emitted by the color center. This promising result could be improved by integrating the emitters into photonic structures designed to increase collection and reduce the impact of dephasing.The results detailed in this thesis show the potential of this family of blue color centers in hBN as quantum emitters for quantum information applications. With future developments aiming to understand and control their dynamics, as well as to integrate them into optoelectronic devices, our work opens new perspectives for optical quantum information with 2D materials

The notion that a text is coherent in virtue of the `relations' that hold between its component spans currently forms the basis for an active research programme in discourse linguistics. Coherence relations feature prominently in many theories of discourse structure, and have recently been used with considerable success in text generation systems. However, while the concept of coherence relations is now common currency for discourse theorists there remains much confusion about them, and no standard set of relations has yet emerged. The aim of this thesis is to contribute towards the development of a standard set of relations. We begin from an explicitly empirical conception of relations: they are taken to model a collection of psychological mechanisms operative during the tasks of reading and writing. This conception is fleshed out with reference to psychological theories of skilled task performance, and to Rosch's notion of the basic level of categorisation. A methodology for investigating these mechanisms is then presented, which takes as its starting point a study of cue phrases- the sentence/clause connectives by which they are signaled. Although it is conventional to investigate psychological mechanisms by studying human behaviour, it is argued here that evidence for the constructs modelled by relations can be sought in ananalysis of the linguistic resources available for marking them explicitly intext. The methodology is based on two simple linguistic tests: the test for cue phrases and the test for substitutability. Both tests are functional in inspiration: the former test identifies a heterogenous class of phrases used for linking one portion of text to another; and the later test is used to discover when a writer is willing to substitute one of these phrases for another. The tests are designed to capture the judgements of ordinary readers and writers, rather than the theoretical intuitions of specialised discourse analysts. The test for cue phrases is used to analyse around 20 pages of naturally occuring text, from which a corpus of over 20 cue phrases is assembled. The substitutability test is then used to organise this corpus into a hierarchical taxonomy, representing the substitutability relationship between every pair of phrases. The taxonomy of cue phrases lends itself neatly to a model of relations as feature-based constructs. Many cue phrases can be interpreted as signalling just some features of relations, rather than whole relations. Small extracts from the taxonomy can be used systematically to determine the alternative values of single features; complex relation definitions can then be formed by combining the values of many features. The thesis delivers results on two levels. Firstly,it sets out a methodology for motivating a set of relation definitions, which rests on a systematic analysis of oncrete linguistic data, and demands a minimum of theoretical assumptions. Also provided are the relation definitions which result from applying the methodology. The new definitions give an interesting picture of the variation that exists amongst cuephrases, and offers a number of innovative insights into text coherence.

L'étude de la physique des systèmes à l'échelle nanométrique nécessite idéalement une résolution spatiale atomique et une résolution temporelle de l'ordre de la femtoseconde. La microscopie électronique en transmission ultra-rapide (UTEM), qui combine une résolution temporelle inférieure à la picoseconde et une résolution spatiale nanométrique, est récemment apparue comme un outil unique doté de résolutions spatio-temporelles sans précédent. Cependant, les performances des premiers UTEMs étaient limitées par la brillance des photocathodes utilisées comme source d'électrons ultra-rapide. Dans ce contexte, il a été vite réalisé que les UTEMs utilisant des sources d'électrons déclenchées par laser et basées sur des pointes métalliques comme émetteur, permettraient de dépasser cette limitation. L'objectif de cette thèse est de décrire le développement d'un microscope électronique en transmission ultra-rapide basé sur ce type de source dites " source à émission de champ froide ", pouvant fonctionner en mode continu ou ultra-rapide. L'émission d'électrons d'une nanopointe de tungstène est déclenchée par des impulsions laser femtosecondes, qui sont fortement focalisées par des composants optiques insérés proche de la cathode. La brillance mesurée est la plus grande mesurée à ce jour dans un UTEM. En associant cette nouvelle source de brillance élevée à un système d'injection/cathodoluminescence composé d'un miroir parabolique installé au-dessus du porte-échantillon, l'UTEM peut être utilisé pour réaliser des expériences TEM-pompe-sonde ultra-rapides résolues dans le temps. Les possibilités d'un tel instrument pour l'imagerie ultra-rapide, la diffraction, l'holographie électronique et la spectroscopie sont présentées. Une attention particulière a été accordée aux applications en nano-optique. La spectroscopie électronique de gain d'énergie (EEGS) permet notamment d'étudier les excitations optiques de nano-systèmes à travers les modifications du spectre d'énergie des électrons. La possibilité de synchroniser facilement les impulsions d'électrons libres ultra-courts avec l'excitation optique de l'échantillon dans les UTEM est essentielle pour l'observation d'interactions électron/photon fortement non linéaires. Ces expériences nous ont permis de caractériser les propriétés spectro-temporelles du faisceau d'électrons ultra-courts. La dernière partie propose une discussion des premières expériences d'holographie électronique hors-axe réalisées avec des impulsions électroniques ultra-rapides. En effet, la dose d'électrons dans le plan de l'échantillon étant considérablement réduite en raison du faible taux de répétition du train d'impulsions électronique, les hologrammes ultra-rapides sont obtenus dans des conditions dites " low dose " complexes. En conséquence, les paramètres expérimentaux couramment utilisés pour l'acquisition d'hologrammes avec des TEM conventionnels ne peuvent pas être directement utilisés en mode ultra-rapide. Des études expérimentales ont été réalisées pour déterminer les conditions optimales pour l'holographie électronique hors-axe ultra-rapide. L'influence de la dose, la longueur de cohérence de la source, les conditions d'illumination et les instabilités de l'instrument ont été systématiquement prises en compte
The investigation of the physics of nanoscale systems ideally requires atomic spatial resolution and femtosecond time-resolution. Ultrafast Transmission Electron Microscopy (UTEM) combining subpicosecond temporal resolution and nanometer spatial resolution has recently emerged as a unique tool with unprecedented spatio-temporal resolutions. However, the performances of the first UTEMs were limited by the brightness of the photocathodes used as ultrafast electron source. In this context, it was soon realized that UTEMs relying on laser-driven electron sources based on nanoscale emitters would overcome this limitation. The aim of this thesis is to report the development of an ultrafast Transmission Electron Microscope based on a cold field emission source, which can operate either in DC or ultrafast mode. Electron emission from a tungsten nanotip is triggered by femtosecond laser pulses, which are tightly focused by optical components integrated inside a cold-field emission source close to the cathode. The measured brightness is the largest reported so far for UTEMs. Combining this new high brightness source with an injection/Cathodoluminescence system, composed of a parabolic mirror placed above the sample holder, the UTEM can be used to perform time-resolved ultrafast pump-probe TEM experiments. The possibilities of such an instrument for ultrafast imaging, diffraction, electron holography and spectroscopy are presented. Particular attention has been paid on applications in nano-optics. In particular, Electron Energy Gain Spectroscopy (EEGS) allows to investigate the optical excitations of nanosystems in the energy domain. The ability to easily synchronize ultrashort free electron pulses with the optical excitation of the sample in UTEMs is essential for the observation of strongly nonlinear electron/photon interactions. These experiments will enable us to characterize the spectro-temporal properties of the ultrashort electron beam. Off-axis electron holography performed with ultrafast electron pulses are finally discussed. The electron dose in the specimen plane is considerably reduced due to the low repetition rate of the electron pulse train. This peculiar property of ultrafast FE-TEMs implies that ultrafast holograms are acquired in low-dose-like conditions. As a consequence, the experimental parameters commonly used for the acquisition of off-axis electron holograms with conventional TEMs cannot be directly implemented in the ultrafast mode. Experimental studies were performed to find the optimum conditions for ultrafast off-axis electron holography. Influence of the dose, the coherence length of the source, the illumination condition and the instrument instabilities have been addressed

Assessing executive attention in human is of paramount importance in understanding the development of critical neural pathways of the brain. In particular, it deals with one’s ability to monitor and resolve cognitive conflicts. In the literature, the efficiency of the executive attention is widely assessed based on the conflict-related changes in the behavioural data. In this thesis, we introduce a new data-driven approach, which exploits Electroencephalography (EEG) coherency to study the topography and efficiency of the executive attention.

This report quantifies the motion of wheat subject to wind and assesses the effect of this motion on the coherence obtained from Synthetic Aperture Radar (SAR) interferometry. Over vegetation, the loss of coherence due to the change in backscatter between two SAR images taken at a different time (temporal decorrelation) is related to the wind induced motion of vegetation elements. The research aims to provide simultaneous in situ measurements of crop motion and wind velocity at canopy height and to use these measurements in a coherence model to determine the quantitatively the parameters which infer temporal decorrelation. The potential of coherence for agricultural applications is assessed. The three-dimensional motion of wheat is measured by a photogrammetry method using two commercially available video cameras. Simultaneously, wind velocity at canopy height is measured by anemometers at a high sampling frequency. Wheat motion and wind velocity data were collected in a field local to Cranfield University in summer 2000. The CD attached to this report contains the wheat motion and wind velocity data. They show that the motion of wheat is correlated with the wind speed, and that wheat plants adjacent to each other move coherently. The coherence model is based on a statistical approach, which represents the total backscatter from vegetation as the phasor addition of a fixed component and one or more components which are weather dependent. The relative contributions of the total backscatter are estimated with the RT2 backscatter intensity model. The motion measurements are used to define the variability of the phase of the weather dependent components in the model. Outputs of the model show that a C-band SAR with an incidence angle of 23° (typical configuration of the ERS satellites) yields coherence values highly variable with the wind conditions at the time of the radar passes. The potential use of coherence for agricultural applications is limited by this variability, which infers the need for an accurate coherent backscatter model.

Thesis (M.S.) -- University of Maryland, College Park, 2006.
Thesis research directed by: Dept. of Electrical and Computer Engineering. Title from t.p. of PDF. Includes bibliographical references. Published by UMI Dissertation Services, Ann Arbor, Mich. Also available in paper.

Des sources des rayons XUV (1-100 nm) sont des outils extraordinaires pour sonder la dynamique à l’échelle nanométrique avec une résolution femto- voire attoseconde. La génération d’harmoniques d’ordre élevé (GH) est une des sources majeures dans ce domaine d’application. La GH est un processus dans lequel une impulsion laser infrarouge femtoseconde est convertie, de manière cohérente, en fréquences élevées dans le domaine EUV par interaction hautement non-linéaire dans un atome, une molécule et plus récemment, dans un cristal. La GH possède une excellente cohérence spatiale qui a permis de réaliser des démonstrations impressionnantes en imagerie sans lentille. Pour accroître le potentiel de ces sources, des défis sont à relever : leur brillance et énergie de photon maximum doivent augmenter et les techniques d’imagerie sans lentille doivent être modifiées pour être compatibles avec l’importante largeur spectrale des impulsions attosecondes émise par ces sources. Cette thèse présente une nouvelle approche dans laquelle des figures de diffraction large bande, i.e. potentiellement attosecondes, sont rendues monochromatiques numériquement. Cette méthode est basée uniquement sur la mesure du spectre de la source et la supposition d’un échantillon spatialement non-dispersif. Cette approche a été validée tout d’abord dans le visible, à partir d’un supercontinuum. L’échantillon binaire est reconstruit par recouvrement de phase pour une largeur spectrale de 11 %, là où les algorithmes usuels divergent. Les simulations numériques montrent aussi que la méthode de monochromatisation peut être appliquée au domaine des rayons X, avec comme exemple un masque semi-conducteur utilisé en de lithographie EUV. Bien que la brillance « cohérente » de la source actuelle (qui progresse) reste insuffisante, une application sur l’inspection de masques sur source Compton est proposée. Dans une extension de ces simulations un masque de lithographie étendu est reconstruit par ptychographie, démontrant la versatilité à d’autres techniques d’imagerie sans lentille. Nous avons également entamé une série d’expérience dans le domaine des X-durs sur source synchrotron. Les figures de diffraction après monochromatisation numérique semblent prometteuses mais l’analyse des données demandent des efforts supplémentaires. Une partie importante de cette thèse est dédiée à l’extension des sources harmoniques à des brillances et énergies de photon plus élevées. Ce travail exploratoire permettrait la réalisation d’une source harmonique compacte pompée par un laser OPCPA dans le moyen infrarouge à très fort taux de répétition. Les longueurs d’onde moyen infrarouge (3.1 μm dans ce travail de thèse) sont favorables à l’extension des énergies des photons au keV et aux impulsions attosecondes. Le but est de pouvoir couvrir les seuils d’absorption X et d’améliorer la résolution spatio-temporelle. Cependant, deux facteurs rendent cette démonstration difficile: le nombre de photons par impulsion de la source OPCPA est très limité et la réponse du dipôle harmonique à grande longueur est extrêmement faible. Pour relever ces défis plusieurs configurations expérimentales sont explorées : génération dans un jet de gaz ; génération dans une cellule de gaz ; compression solitonique et la génération d’harmoniques combinées dans une fibre à cristal photonique ; compression solitonique dans une fibre à cristal photonique et génération d’harmoniques dans une cellule de gaz. Les premiers résultats expérimentaux sur la compression solitonique jusqu’à 26 femtosecondes et des harmoniques basses jusqu’à l’ordre sept sont présentésEn résumé, ces résultats représentent une avancée vers l’imagerie nanométrique attoseconde sans lentille basée sur des algorithmes « large bande » innovants et une extension des capacités de nouvelles sources harmoniques ‘table-top’ au keV pompées par laser OPCPA
Soft X-ray sources based on high harmonic generation are up to now unique tools to probe dynamics in matter on femto- to attosecond timescales. High harmonic generation is a process in which an intense femtosecond laser pulse is frequency upconverted to the UV and soft X-ray region through a highly nonlinear interaction in a gas. Thanks to their excellent spatial coherence, they can be used for lensless imaging, which has already led to impressive results. To use these sources to the fullest of their potential, a number of challenges needs to be met: their brightness and maximum photon energy need to be increased and the lensless imaging techniques need to be modified to cope with the large bandwidth of these sources. For the latter, a novel approach is presented, in which broadband diffraction patterns are rendered monochromatic through a numerical treatment based solely on the spectrum and the assumption of a spatially non-dispersive sample. This approach is validated through a broadband lensless imaging experiment on a supercontinuum source in the visible, in which a binary sample was properly reconstructed through phase retrieval for a source bandwidth of 11 %. Through simulations, the numerical monochromatization method is shown to work for hard X-rays as well, with a simplified semiconductor lithography mask as sample. A potential application of lithography mask inspection on an inverse Compton scattering source is proposed, although the conclusion of the analysis is that the current source lacks brightness for the proposal to be realistic. Simulations with sufficient brightness show that the sample is well reconstructed up to 10 % spectral bandwidth at 8 keV. In an extension of these simulations, an extended lithography mask sample is reconstructed through ptychography, showing that the monochromatization method can be applied in combination with different lensless imaging techniques. Through two synchrotron experiments an experimental validation with hard X-rays was attempted, of which the resulting diffraction patterns after numerical monochromatization look promising. The phase retrieval process and data treatment however require additional efforts.An important part of the thesis is dedicated to the extension of high harmonic sources to higher photon energies and increased brightness. This exploratory work is performed towards the realization of a compact high harmonic source on a high repetition rate mid-IR OPCPA laser system, which sustains higher average power and longer wavelengths compared to ubiquitous Ti:Sapphire laser systems. High repetition rates are desirable for numerous applications involving the study of rare events. The use of mid-IR wavelengths (3.1 μm in this work) promises extension of the generated photon energies to the kilo-electronvolt level, allowing shorter pulses, covering more X-ray absorption edges and improving the attainable spatial resolution for imaging. However, high repetition rates come with low pulse energies, which constrains the generation process. The generation with longer wavelengths is challenging due to the significantly lower dipole response of the gas. To cope with these challenges a number of experimental configurations is explored theoretically and experimentally: free-focusing in a gas-jet; free-focusing in a gas cell; soliton compression and high harmonic generation combined in a photonic crystal fiber; separated soliton compression in a photonic crystal fiber and high harmonic generation in a gas cell. First results on soliton compression down to 26 fs and lower harmonics up to the seventh order are presented.Together, these results represent a step towards ultrafast lensless X-ray imaging on table-top sources and towards an extension of the capabilities of these sources

Dans ce travail de thèse, je présente une série d'études expérimentales réalisées avec des gaz d'ytterbium ultrafroids excités sur différentes transitions optiques. L'ytterbium appartient à la famille des atomes dits alcalino-terreux. Ces atomes possèdent une structure électronique riche, avec une transition d’horloge exempte d’émission spontanée et une transition étroite d'intercombinaison. Avec des expériences de spectroscopie sur la transition d'horloge dans des réseaux optiques profonds, je montre notre capacité à exciter cette transition de manière cohérente pendant de longues périodes. Ce contrôle est ensuite utilisé dans un premier temps en tant qu'outil pour mesurer les longueurs de diffusion des états impliqués dans la transition d'horloge et ensuite, pour préparer un petit système quantique ouvert, où la dissipation prend la forme de pertes à deux corps. En branchant ce couplage adiabatiquement, on observe une forte suppression de ces pertes, ce qui est interprété comme une signature de l'effet Zénon quantique. J'utilise finalement ce transfert cohérent pour étudier la dynamique de relaxation d'un condensat de Bose-Einstein. Enfin, je développe une étude sur un système ouvert avec des interactions fortes. Ici, on induit artificiellement de la dissipation sous la forme d'émission spontanée en utilisant la transition d'intercombinaison, et j'étudie comment la cohérence spatiale d'un superfluide dans un réseau optique est détruite. Ces expériences révèlent que la présence d'interactions fortes protège partiellement une cohérence résiduelle et entraîne un développement non-triviale de la décohérence, révélant ainsi l'émergence d'un canal de relaxation subdiffusif
In this thesis I report on a series of experimental studies performed with ultracold ytterbium gases driven in different optical transitions. Ytterbium belongs to the family of the so-called alkaline-earth-like atoms, which feature a rich electronic structure, with an optical clock transition free of spontaneous emission, and a narrow intercombination transition, making them very appealing for metrological and quantum simulation proposals. By performing spectroscopy on the clock transition, I prove on a first set of experiments in deep optical lattices our ability to drive this transition coherently for long times. This coherent control is then used for different studies. First, as tool to measure the scattering lengths of the states involved in the clock transition. Then, to prepare a small open quantum system, where dissipation arises in the form of two-body losses. By enabling the coupling adiabatically, we observe a strong suppression of these losses, which is interpreted as a signature of the quantum Zeno effect. I ultimately use the coherent driving to study the relaxation dynamics of a dissipative bulk Bose-Einstein condensate. Finally, I elaborate an investigation on a strongly-interacting open system. Dissipation is artificially induced in the form of spontaneous emission using the intercombination transition. Here, I study in which manner spontaneous emission destroys the spatial coherence of a superfluid in an optical lattice. These experiments reveal that the presence of strong interactions partially protects a residual amount of coherence and makes decoherence develop in a non-trivial manner, unveiling the emergence of a subdiffusive relaxation channel

La détection sans fil a évolué depuis la découverte de la détection radar en 1886. L’analyse des réflexions électromagnétiques d’objets a ouvert la voie à un large éventail d’applications allant de la localisation de cibles à longue distance pour la navigation civile et militaire à la surveillance du vent et des précipitations les prévisions météo à la détection de vitesse pour la sécurité routière. Cependant, pendant très longtemps, la détection sans fil a rarement été utilisée pour des applications centrées sur l’homme en raison de limitations techniques, d’impraticabilité ou de coût. L’introduction des réseaux sans fil a suscité́ un nouvel intérêt pour le développement de nouveaux services de détection sans fil en raison de leur souplesse et de leur polyvalence. L’intégration de ces fonctionnalités contribuerait à résoudre certains problèmes de société importants. La localisation, la détection de mouvements et la surveillance des signes vitaux ont un grand potentiel pour promouvoir le vieillissement en bonne santé, la sécurité publique et le commerce. La détection sans contact offre un degré de liberté appréciable, permettant de surveiller à distance les personnes âgées isolées sans entraver leur vie quotidienne. Elle pourrait aider les services de sécurité publique à dénombrer les foules et à détecter les survivants à l’intérieur des bâtiments en cas d’urgence. Les commerces de détail et les établissements publics tireraient parti d’une localisation active et passive pour offrir une expérience améliorée à leurs visiteurs et faciliter leurs efforts logistiques. Cette thèse aborde le problème de l’exploitation des réseaux sans fil commerciaux pour les applications de détection : L’un des défis de la surveillance sans fil consiste à détecter l’orientation d’une personne avec précision. Tandis que d’autres travaux fournissent des solutions à granularité grossière pour résoudre de tels problèmes, nous utilisons les techniques de radar MIMO pour fournir un système d’estimation d’orientation précis pour lesinfrastructures Wi-Fi. Plus précisément, nous analysons les informations de phase des signaux reçus sur le réseau d’antennes afin de calculer le cap d’un terminal Wi-Fi. Un deuxième défi consiste à fournir un système de positionnement précis aux systèmes LPWAN afin de maintenir la cohérence des informations des capteurs déployés. Les solutions actuelles sont complexes, coûteuses ou consomment beaucoup d’énergie. Pour résoudre ce problème, nous introduisons les fonctions MIMO dans les systèmes LoRa LPWAN afin de permettre une localisation précise avec des coûts de démarrage limités. Nous activons l’estimation de l’angle d’arrivée en utilisant une deuxième antenne sur la passerelle LoRaWAN. Nous prouvons également l’utilité de ces informations pour augmenter l’efficacité des communications sans fil. Un troisième défi pour la localisation sans fil est l’inefficacité des approches actuelles basées sur un modèle en cas de conditions de non-visibilité et la rigidité des approches basées sur les données en cas de changements d’environnement de propagation. Pour relever ce défi, nous proposons une nouvelle solution de localisation passive pilotée par les données afin de remédier aux limitations des techniques de localisation basées sur un modèle. Pour donner vie à de tels systèmes et leur donner une chance d’impact sur notre quotidien, nous devons promouvoir la réutilisabilité et la reproductibilité. Pour cela, nous essayons de relever le défi de la reproductibilité dans les réseaux sans fil enanalysant l’état actuel, en réalisant une étude de cas et en présentant les enseignements qui en découlent
Wireless sensing has evolved since the discovery of radio wave echo detection and radar in 1886. Analyzing electromagnetic reflections from objects opened the way for a wide range of applications spanning from locating long-range targets for navigation and military to monitoring wind and precipitation for weather-forecasting to velocity detection for public safety. However, for the longest time, its usefulness was seldom for human-centric applications because of technical limitations, impracticality or costliness. Introducing wireless networks awakened a newfound interest in developing new wireless sensing services for their seamlessness and versatility. Integrating such functionalities would contribute to resolving some prominent societal issues. Localization, motion detection, and vital signs monitoring have great potential for promoting healthy aging, public safety, and retail. Contactless sensing offers an appreciable degree of freedom, enabling remote monitoring of the isolated elderly without hampering their daily lives. It could assist public safety services for crowd counting and detection of survivors inside buildings during emergencies. Retail and public facilities would benefit from passive and active localization to offer an enhanced experience to their visitors and to help their logistical efforts. This thesis addresses the problem of leveraging commercial off-the-shelf wireless networks for sensing applications: One challenge for wireless monitoring is to detect the attitude of a person accurately. While other works provide coarse-grained solutions for resolving such issues, we use MIMO radar techniques to provide an accurate orientation estimation system for Wi-Fi infrastructures. To be more precise, we analyze the phase information of signals received on the antenna array to compute the heading of a Wi-Fi terminal. A second challenge is to provide an accurate positioning system for LPWAN systems to maintain the information consistency of deployed sensors. Current solutions are complex, costly, or not energy-efficient. To address this problem, we introduce MIMO capabilities to LoRa LPWAN systems that provide accurate localization with limited startup costs. We enable the angle of arrival estimation by leveraging a second antenna on the LoRaWAN gateway. We also prove the usefulness of such information for wireless communication efficiency. A third challenge for wireless localization is the inefficiency of current model-based approaches in case of non-line-of-sight conditions and the rigidity of data-driven approaches in case of propagation environment changes. To address this challenge, we propose a new data-driven solution for passive localization to address the limitations of model-based localization techniques. To give life to such systems and provide them with a chance of impacting our everyday lives, we should promote reusability and reproducibility. For that, we focus on the challenge of reproducibility in wireless networking by surveying the current state, performing a case study, and presenting the engendered lessons

La localisation d'une personne ou d'un bien dans des environnements intérieurs est devenue une nécessité. Le système de positionnement par satellites, une solution prédominante pour la localisation en extérieur, rencontre des limites lorsqu'il est appliqué en intérieur en raison de la réflexion des signaux et de l'atténuation causée par les obstacles. Pour y remédier, diverses solutions de localisation en intérieur ont été étudiées. Les méthodes de localisation en intérieur sans fil exploitent les signaux pour déterminer la position d'un appareil dans un environnement intérieur. Cependant, l'interférence des signaux, souvent causée par des obstacles physiques, des réflexions et des appareils concurrents, peut entraîner des imprécisions dans l'estimation de la position. De plus, ces méthodes nécessitent le déploiement d'infrastructures, ce qui entraîne des coûts d'installation et de maintenance. Une autre approche consiste à estimer le mouvement de l'utilisateur à l'aide des capteurs inertiels de l'appareil. Toutefois, cette méthode se heurte à des difficultés liées à la précision des capteurs, aux caractéristiques de mouvement de l'utilisateur et à la dérive temporelle. D'autres techniques de localisation en intérieur exploitent les champs magnétiques générés par la Terre et les structures métalliques. Ces techniques dépendent des appareils et des capteurs utilisés ainsi que de l'environnement dans lequel se situe l'utilisateur.L'objectif de cette thèse est de réaliser un système de localisation en intérieur conçu pour les professionnels, tels que les pompiers, les officiers de police et les travailleurs isolés, qui ont besoin de solutions de positionnement précises et robustes dans des environnements intérieurs complexes. Dans cette thèse, nous proposons un système de localisation en intérieur qui exploite les récentes avancées en vision par ordinateur pour localiser une personne à l’intérieur d’un bâtiment. Nous développons un système de localisation au niveau de la pièce. Ce système est basé sur l'apprentissage profond et les capteurs intégrés dans le smartphone, combinant ainsi les informations visuelles avec le cap magnétique du smartphone. Pour se localiser, l'utilisateur capture une image de l'environnement intérieur à l'aide d'un smartphone équipé d'une caméra, d'un accéléromètre et d'un magnétomètre. L'image capturée est ensuite traitée par notre système composé de plusieurs réseaux neuronaux convolutionnels directionnels pour identifier la pièce spécifique dans laquelle se situe l'utilisateur. Le système proposé nécessite une infrastructure minimale et fournit une localisation précise. Nous soulignons l'importance de la maintenance continue du système de localisation en intérieur par vision. Ce système nécessite une maintenance régulière afin de s'adapter à l'évolution des environnements intérieurs, en particulier lorsque de nouvelles pièces doivent être intégrées dans le système de localisation existant. L'apprentissage incrémental par classe est une approche de vision par ordinateur qui permet aux réseaux neuronaux profonds d'intégrer de nouvelles classes au fil du temps sans oublier les connaissances déjà acquises. Dans le contexte de la localisation en intérieur par vision, ce concept doit être appliqué pour prendre en compte de nouvelles pièces. La sélection d'échantillons représentatifs est essentielle pour contrôler les limites de la mémoire, éviter l'oubli et conserver les connaissances des classes déjà apprises. Nous développons une méthode de sélection d'échantillons basée sur la cohérence pour l'apprentissage incrémental par classe dans le cadre de l'apprentissage profond. La pertinence de la méthodologie et des contributions algorithmiques de cette thèse est rigoureusement testée et validée par des expérimentations et des évaluations complètes sur des données réelles
The need to determine the location of individuals or objects in indoor environments has become an essential requirement. The Global Navigation Satellite System, a predominant outdoor localization solution, encounters limitations when applied indoors due to signal reflections and attenuation caused by obstacles. To address this, various indoor localization solutions have been explored. Wireless-based indoor localization methods exploit wireless signals to determine a device's indoor location. However, signal interference, often caused by physical obstructions, reflections, and competing devices, can lead to inaccuracies in location estimation. Additionally, these methods require access points deployment, incurring associated costs and maintenance efforts. An alternative approach is dead reckoning, which estimates a user's movement using a device's inertial sensors. However, this method faces challenges related to sensor accuracy, user characteristics, and temporal drift. Other indoor localization techniques exploit magnetic fields generated by the Earth and metal structures. These techniques depend on the used devices and sensors as well as the user's surroundings.The goal of this thesis is to provide an indoor localization system designed for professionals, such as firefighters, police officers, and lone workers, who require precise and robust positioning solutions in challenging indoor environments. In this thesis, we propose a vision-based indoor localization system that leverages recent advances in computer vision to determine the location of a person within indoor spaces. We develop a room-level indoor localization system based on Deep Learning (DL) and built-in smartphone sensors combining visual information with smartphone magnetic heading. To achieve localization, the user captures an image of the indoor surroundings using a smartphone, equipped with a camera, an accelerometer, and a magnetometer. The captured image is then processed using our proposed multiple direction-driven Convolutional Neural Networks to accurately predict the specific indoor room. The proposed system requires minimal infrastructure and provides accurate localization. In addition, we highlight the importance of ongoing maintenance of the vision-based indoor localization system. This system necessitates regular maintenance to adapt to changing indoor environments, particularly when new rooms have to be integrated into the existing localization framework. Class-Incremental Learning (Class-IL) is a computer vision approach that allows deep neural networks to incorporate new classes over time without forgetting the knowledge previously learned. In the context of vision-based indoor localization, this concept must be applied to accommodate new rooms. The selection of representative samples is essential to control memory limits, avoid forgetting, and retain knowledge from previous classes. We develop a coherence-based sample selection method for Class-IL, bringing forward the advantages of the coherence measure to a DL framework. The relevance of the methodology and algorithmic contributions of this thesis is rigorously tested and validated through comprehensive experimentation and evaluations on real datasets

博士
國立交通大學
電子物理系
88
We have studied a mesoscopic ring threaded by a magnetic flux that increases linearly with time. The ring is partially coherent such that conduction electrons in the ring will encounter incoherent scatterings. Besides, we have also considered the behavior of the ring when the electrons encounter elastic scatterings due to the presence of an impurity in the ring. We have adopted a S-matrix model, as proposed by Buttiker, for the incoherent scatterings in this time-dependent situation. This allows us to treat the incoherent scatterings, the elastic scatterings and the coherent inelastic processes on the same footing. We have solved exactly the problem that the coherent inelastic processes caused by the time-varying magnetic flux. Our results demonstrate unequivocally that, 1.When there is no impurity, for the electrons emanating out of incoherent scatterings, the lower the energies of these electrons, the greater will be their net contribution to the dc component of the induced current. 2. When there is an impurity : (a)In the case of a weak impurity, the lower the energies of the electrons that emanate out of incoherent scatterings, the greater will be their net contribution to the dc component I_{dc} of the induced current just like the situation of no impurity. (b)In the case of a strong impurity, however, I_{dc} alternates between regions of zero and nonzero values as the chemical potential \mu increases. The peak value of I_{dc} in the nonzero region increases with \mu. We find that these regions of zero I_{dc}, and nonzero I_{dc}, correspond closely with the gaps, and the bands, respectively, of a one dimensional energy band. These characteristics arise from the fact that the electrons traversing the ring have their {\it energies} shifted gradually until their {\it energies} fall upon a forbidden region, where they suffer total reflection. This kind of total reflection does not occur in a ring with a constant flux. Our results thus contain the nonadiabatic effects of the changing flux on the dissipation in a partially coherent ring. The evolution of these nonadiabatic features in the intermediate impurity regime is also investigated.

An appropriate design is required to reduce the problems associated with the software development and maximize the benefits. Modularization in the other way can help in improving the design which can also help in Global Software Development(GSD). In GSD, Modularization can help in distributing the work in different regions. Work is distributed after a thorough study of different tasks and various location, So to do this, categorization of work should be undertaken before it is distributed. This thesis is overall divided into two parts. The first part helps in finding modules using two different analysis. The second part uses the final output of the first part to find coupling and cohesion. In globally distributed software projects,it is possible to have a discrepancy between organizational structure and software architecture which may diminish the overall production. The final aim of software modularization is to increase software project quality and productivity. This thesis is subdivided into two different parts. The first part helps in finding modules using two different analysis : Dependency and Functionality Analysis. In dependency analysis the dependency among modules are checked and the best modular structure is found of it and in functionality analysis the components having functions together are taken to form the best modular structure and after this the best modular structure is found out using both the analysis. In the second part, coupling and cohesion are found using information theory approach .

碩士
國立臺灣大學
工程科學及海洋工程學研究所
106
Under high wind conditions, the surface layer is characterized by an elongated flow structure which manifests itself by forming streaks of bubbles and droplets on the surface. Despite intermittent disruption by breaking waves, the persistent and structural features of these surface streaks suggest the existence of underlying coherent vortices that form the surface streaks and the self-sustaining mechanisms of the vortices. High-resolution numerical simulation of a turbulent flow subject to high shear was conducted to elucidate the coherent vortical structures and their formation mechanism. The flow is driven by imposing shear stress (surface friction velocity on the water side = 1.225 cm/s) on the boundary to mimic the surface layer. For comparison, computations were also conducted for the surface layer with weaker shear rates (with surface friction velocity = 0.707 and 1 cm/s). The vortical structures were identified by adopting an indicator of swirling strength derived from the local velocity gradient tensor. These vortical structures are marked with two different colors to distinguish the rotation direction and the vortical structures responsible for various flow processes and surface signatures can be extracted by employing the advanced variable-interval space-averaging technique. The simulated shear flow reveals two distinct surface signatures: elongated high-speed streaks that are arranged with somewhat equal cross-spacing and localized low-speed spots that appear intermittently. Three types of vortical structures were identified as responsible for these surface signatures: forward horseshoe vortices (head heading downstream) associated with upwelling and surface divergence that can cause surface low-speed spots; reversed horseshoe vortices (head heading upstream) associated with downwelling that can strengthen surface high-speed streaks; quasi-streamwise vortices that flank high-speed streaks as staggered, counter-rotating arrays. This vortical structure supports the regeneration mechanism in which pre-existing parent vortices interact with the sheared surface and induce offspring vortices. Surface layers with different shear rates exhibit similar non-dimensional streak spacing in a viscous scale; however, the streamwise extent, increases with the shear rate, indicating prolonged surface streaks under high-wind conditions.

Typical ultrafast plane-wave ultrasound imaging involves: 1) insonifying the medium with several plane-wave pulses emitted at different angles by a linear transducer array, 2) sampling the returning echo signals, after each plane-wave emission, with the same transducer array, 3) beamforming the recorded angle-specific raw data frames, and 4) compounding the beamformed data frames over all angles to form a final image. This thesis attempts to address the following question: Given a set of available plane-wave emission angles, which ones should we select for acquisition (i.e., which angle-specific raw data frames should we sample), to achieve adequate image quality at low cost associated with both sampling and computation? We propose a simple similarity-driven angle selection scheme and evaluate its several variants that rely on user-specified similarity measurement thresholds guiding the recursive angle selection process. Our results show that the proposed scheme has a low computational overhead and can yield significant savings in terms of the amount of sampled raw data.
Graduate

碩士
國立清華大學
資訊工程學系
100
Chip Multiprocessor(CMP) is becoming the norm of processor chips. To design CMP, tracedriven simulation has been a commonly used technique for fast exploration of architecture design space. With the availability of parallel computers, such as Tilera’s Tile64, parallel trace-driven simulation for faster architecture evaluation is becoming possible. However, there are very few papers discussing parallel trace-driven simulation. This thesis discusses the design and implementation of a parallel trace-driven simulator for NoC-based cache coherence CMP named TileSim+, TileSim+ provides cycle-accurate network model and cycle-count accurate cache simulation model, which allows the precise evaluation of memory access delay but exploration of cache design space for NoC-based CMP. Most importantly, accelerated with machine such as Tilera’s Tile64, TileSim+ speeds up trace-driven simulation with good scalability. The experimental evaluation of TileSim+ on TILE64 shows that it can obtain correct simulation results for the tested benchmark programs and achieve good speedup over sequential simulator. We also demonstrate how to use TileSim+ to evaluate CMP cache designs.

Realistic computer generated images are computed by combining geometric effects, reflectance models for several captured and phenomenological materials, and real-world lighting according to mathematical models of physical light transport. Several important lighting phenomena should be considered when targeting realistic image simulation. A combination of soft and hard shadows, which arise from the interaction of surface and light geometries, provide necessary shape perception cues for a viewer. A wide variety of realistic materials, from physically-captured reflectance datasets to empirically designed mathematical models, modulate the virtual surface appearances in a manner that can further dissuade a viewer from considering the possibility of computational image synthesis over that of reality. Lastly, in many important cases, light reflects off many different surfaces before entering the eye. These secondary effects can be critical in grounding the viewer in a virtual world, since the human visual system is adapted to the physical world, where such effects are constantly in play. Simulating each of these effects is challenging due to their individual underlying complexity. The net complexity is compounded when several effects are combined. This thesis will investigate real-time approaches for simulating these effects under stringent performance and memory constraints, and with varying degrees of interactivity. In order to make these computations tractable given these added constraints, I will use data and signal analysis techniques to identify predictable patterns in the different spatial and angular signals used during image synthesis. The results of this analysis will be exploited with several analytic and data-driven mathematical models that are both efficient, and yield accurate approximations with predictable and controllable error.

In recent times there has been an explosion of online user-generated video content. This has generated significant research interest in video analytics. Human users understand videos based on high-level semantic concepts. However, most of the current research in video analytics are driven by low-level features and descriptors, which often lack semantic interpretation. Existing attempts in semantic video analytics are specialized and require additional resources like movie scripts, which are not available for most user-generated videos. There are no general purpose approaches to understanding videos through semantic concepts. In this thesis we attempt to bridge this gap. We view videos as collections of entities which are semantic visual concepts like the persons in a movie, or cars in a F1 race video. We focus on two fundamental tasks in Video Understanding, namely summarization and scene- discovery. Entity-driven Video Summarization and Entity-driven Scene discovery are important open problems. They are challenging due to the spatio-temporal nature of videos, and also due to lack of apriori information about entities. We use Bayesian nonparametric methods to solve these problems. In the absence of external resources like scripts we utilize fundamental structural properties like temporal coherence in videos- which means that adjacent frames should contain the same set of entities and have similar visual features. There have been no focussed attempts to model this important property. This thesis makes several contributions in Computer Vision and Bayesian nonparametrics by addressing Entity-driven Video Understanding through temporal coherence modeling. Temporal Coherence in videos is observed across its frames at the level of features/descriptors, as also at semantic level. We start with an attempt to model TC at the level of features/descriptors. A tracklet is a spatio-temporal fragment of a video- a set of spatial regions in a short sequence (5-20) of consecutive frames, each of which enclose a particular entity. We attempt to find a representation of tracklets to aid tracking of entities. We explore region descriptors like Covari- ance Matrices of spatial features in individual frames. Due to temporal coherence, such matrices from corresponding spatial regions in successive frames have nearly identical eigenvectors. We utilize this property to model a tracklet using a covariance matrix, and use it for region-based entity tracking. We propose a new method to estimate such a matrix. Our method is found to be much more efficient and effective than alternative covariance-based methods for entity tracking. Next, we move to modeling temporal coherence at a semantic level, with special emphasis on videos of movies and TV-series episodes. Each tracklet is associated with an entity (say a particular person). Spatio-temporally close but non-overlapping tracklets are likely to belong to the same entity, while tracklets that overlap in time can never belong to the same entity. Our aim is to cluster the tracklets based on the entities associated with them, with the goal of discovering the entities in a video along with all their occurrences. We argue that Bayesian Nonparametrics is the most convenient way for this task. We propose a temporally coherent version of Chinese Restaurant Process (TC-CRP) that can encode such constraints easily, and results in discovery of pure clusters of tracklets, and also filter out tracklets resulting from false detections. TC-CRP shows excellent performance on person discovery from TV-series videos. We also discuss semantic video summarization, based on entity discovery. Next, we consider entity-driven temporal segmentation of a video into scenes, where each scene is characterized by the entities present in it. This is a novel application, as existing work on temporal segmentation have focussed on low-level features of frames, rather than entities. We propose EntScene: a generative model for videos based on entities and scenes, and propose an inference algorithm based on Blocked Gibbs Sampling, for simultaneous entity discovery and scene discovery. We compare it to alternative inference algorithms, and show significant improvements in terms of segmentatio and scene discovery. Video representation by low-rank matrix has gained popularity recently, and has been used for various tasks in Computer Vision. In such a representation, each column corresponds to a frame or a single detection. Such matrices are likely to have contiguous sets of identical columns due to temporal coherence, and hence they should be low-rank. However, we discover that none of the existing low-rank matrix recovery algorithms are able to preserve such structures. We study regularizers to encourage these structures for low-rank matrix recovery through convex optimization, but note that TC-CRP-like Bayesian modeling is better for enforcing them. We then focus our attention on modeling temporal coherence in hierarchically grouped sequential data, such as word-tokens grouped into sentences, paragraphs, documents etc in a text corpus. We attempt Bayesian modeling for such data, with application to multi-layer segmentation. We first make a detailed study of existing models for such data. We present a taxonomy for such models called Degree-of-Sharing (DoS), based on how various mixture components are shared by the groups of data in these models. We come up with Layered Dirichlet Process which generalizes Hierarchical Dirichlet Process to multiple layers, and can also handle sequential information easily through Markovian approach. This is applied to hierarchical co-segmentation of a set of news transcripts- into broad categories (like politics, sports etc) and individual stories. We also propose a explicit-duration (semi-Markov) approach for this purpose, and provide an efficient inference algorithm for this. We also discuss generative processes for distribution matrices, where each column is a probability distribution. For this we discuss an application: to infer the correct answers to questions on online answering forums from opinions provided by different users.

In recent times there has been an explosion of online user-generated video content. This has generated significant research interest in video analytics. Human users understand videos based on high-level semantic concepts. However, most of the current research in video analytics are driven by low-level features and descriptors, which often lack semantic interpretation. Existing attempts in semantic video analytics are specialized and require additional resources like movie scripts, which are not available for most user-generated videos. There are no general purpose approaches to understanding videos through semantic concepts. In this thesis we attempt to bridge this gap. We view videos as collections of entities which are semantic visual concepts like the persons in a movie, or cars in a F1 race video. We focus on two fundamental tasks in Video Understanding, namely summarization and scene- discovery. Entity-driven Video Summarization and Entity-driven Scene discovery are important open problems. They are challenging due to the spatio-temporal nature of videos, and also due to lack of apriori information about entities. We use Bayesian nonparametric methods to solve these problems. In the absence of external resources like scripts we utilize fundamental structural properties like temporal coherence in videos- which means that adjacent frames should contain the same set of entities and have similar visual features. There have been no focussed attempts to model this important property. This thesis makes several contributions in Computer Vision and Bayesian nonparametrics by addressing Entity-driven Video Understanding through temporal coherence modeling. Temporal Coherence in videos is observed across its frames at the level of features/descriptors, as also at semantic level. We start with an attempt to model TC at the level of features/descriptors. A tracklet is a spatio-temporal fragment of a video- a set of spatial regions in a short sequence (5-20) of consecutive frames, each of which enclose a particular entity. We attempt to find a representation of tracklets to aid tracking of entities. We explore region descriptors like Covari- ance Matrices of spatial features in individual frames. Due to temporal coherence, such matrices from corresponding spatial regions in successive frames have nearly identical eigenvectors. We utilize this property to model a tracklet using a covariance matrix, and use it for region-based entity tracking. We propose a new method to estimate such a matrix. Our method is found to be much more efficient and effective than alternative covariance-based methods for entity tracking. Next, we move to modeling temporal coherence at a semantic level, with special emphasis on videos of movies and TV-series episodes. Each tracklet is associated with an entity (say a particular person). Spatio-temporally close but non-overlapping tracklets are likely to belong to the same entity, while tracklets that overlap in time can never belong to the same entity. Our aim is to cluster the tracklets based on the entities associated with them, with the goal of discovering the entities in a video along with all their occurrences. We argue that Bayesian Nonparametrics is the most convenient way for this task. We propose a temporally coherent version of Chinese Restaurant Process (TC-CRP) that can encode such constraints easily, and results in discovery of pure clusters of tracklets, and also filter out tracklets resulting from false detections. TC-CRP shows excellent performance on person discovery from TV-series videos. We also discuss semantic video summarization, based on entity discovery. Next, we consider entity-driven temporal segmentation of a video into scenes, where each scene is characterized by the entities present in it. This is a novel application, as existing work on temporal segmentation have focussed on low-level features of frames, rather than entities. We propose EntScene: a generative model for videos based on entities and scenes, and propose an inference algorithm based on Blocked Gibbs Sampling, for simultaneous entity discovery and scene discovery. We compare it to alternative inference algorithms, and show significant improvements in terms of segmentatio and scene discovery. Video representation by low-rank matrix has gained popularity recently, and has been used for various tasks in Computer Vision. In such a representation, each column corresponds to a frame or a single detection. Such matrices are likely to have contiguous sets of identical columns due to temporal coherence, and hence they should be low-rank. However, we discover that none of the existing low-rank matrix recovery algorithms are able to preserve such structures. We study regularizers to encourage these structures for low-rank matrix recovery through convex optimization, but note that TC-CRP-like Bayesian modeling is better for enforcing them. We then focus our attention on modeling temporal coherence in hierarchically grouped sequential data, such as word-tokens grouped into sentences, paragraphs, documents etc in a text corpus. We attempt Bayesian modeling for such data, with application to multi-layer segmentation. We first make a detailed study of existing models for such data. We present a taxonomy for such models called Degree-of-Sharing (DoS), based on how various mixture components are shared by the groups of data in these models. We come up with Layered Dirichlet Process which generalizes Hierarchical Dirichlet Process to multiple layers, and can also handle sequential information easily through Markovian approach. This is applied to hierarchical co-segmentation of a set of news transcripts- into broad categories (like politics, sports etc) and individual stories. We also propose a explicit-duration (semi-Markov) approach for this purpose, and provide an efficient inference algorithm for this. We also discuss generative processes for distribution matrices, where each column is a probability distribution. For this we discuss an application: to infer the correct answers to questions on online answering forums from opinions provided by different users.

Gravity-driven membrane (GDM) filtration is one of the promising membrane bioreactor (MBR) technologies. It operates at a low pressure by gravity, requiring a minimal energy. Thus, it exhibits a great potential for a decentralized system, conducting household in developing and transition countries. Biofouling is a universal problem in almost all membrane filtration applications, leading to the decrease in flux or the increase in transmembrane pressure depending on different operation mode. Air scoring or regular membrane cleaning has been utilized for fouling mitigation, which requires increased energy consumption as well as complicated operations. Besides, repeating cleaning will trigger the deterioration of membranes and shorten their lifetime, elevating cost expenditures accordingly. In this way, GDM filtration stands out from conventional MBR technologies in a long-term operation with relative stable flux, which has been observed in many studies. The objective of this study was to monitor the biofilm development on a flat sheet membrane submerged in a GDM reactor with constant gravitational pressure. Morphology of biofilm layer in a fixed position was acquired by an in-situ and on-line OCT (optical coherence tomography) scanning at regular intervals for both visual investigation and structure analysis. The calculated thickness and roughness were compared to the variation of flux, fouling resistance and permeate quality, showing expected consistency. At the end of experiment, the morphology of entire membrane surface was scanned and recorded by OCT. Membrane autopsy was carried out for biofilm composition analysis by total organic carbon (TOC) and liquid chromatography with organic carbon detection (LC-OCD). In addition, biomass concentration was obtained by flow cytometer and adenosine tri-phosphate (ATP) method. The data of biofilm components indicated a homogeneous biofilm structure formed after a long-term running of the GDM system, based on the morphology observation by OCT images. The superiority of GDM in both flux maintaining and long-term operation with production of high quality effluent was demonstrated, as well as the suitability of OCT for biofouling monitoring was emphasized.