Rendu d'images en demi-tons par diffusion d'erreur sensible à la structure

Le présent mémoire comprend un survol des principales méthodes de rendu en demi-tons, de l’analog screening à la recherche binaire directe en passant par l’ordered dither, avec une attention particulière pour la diffusion d’erreur. Ces méthodes seront comparées dans la perspective moderne de la sensibilité à la structure. Une nouvelle méthode de rendu en demi-tons par diffusion d’erreur est présentée et soumise à diverses évaluations. La méthode proposée se veut originale, simple, autant à même de préserver le caractère structurel des images que la méthode à l’état de l’art, et plus rapide que cette dernière par deux à trois ordres de magnitude. D’abord, l’image est décomposée en fréquences locales caractéristiques. Puis, le comportement de base de la méthode proposée est donné. Ensuite, un ensemble minutieusement choisi de paramètres permet de modifier ce comportement de façon à épouser les différents caractères fréquentiels locaux. Finalement, une calibration détermine les bons paramètres à associer à chaque fréquence possible. Une fois l’algorithme assemblé, toute image peut être traitée très rapidement : chaque pixel est attaché à une fréquence propre, cette fréquence sert d’indice pour la table de calibration, les paramètres de diffusion appropriés sont récupérés, et la couleur de sortie déterminée pour le pixel contribue en espérance à souligner la structure dont il fait partie.

Asset Pricing in a Production Economy with Chew–Dekel Preferences

In this paper we provide a thorough characterization of the asset returns implied by a simple general equilibrium production economy with Chew–Dekel risk preferences and convex capital adjustment costs. When households display levels of disappointment aversion consistent with the experimental evidence, a version of the model parameterized to match the volatility of output and consumption growth generates unconditional expected asset returns and price of risk in line with the historical data. For the model with Epstein–Zin preferences to generate similar statistics, the relative risk aversion coefficient needs to be about 55, two orders of magnitude higher than the available estimates. We argue that this is not surprising, given the limited risk imposed on agents by a reasonably calibrated stochastic growth model.

Musique populaire au cinéma, pastiches et genres : essai poïétique d'un historien-compositeur-bricoleur

Les fichiers audiovisuels complémentaires à ce mémoire sont disponibles sur DVD. Un
 de
 ces

 DVD
 sera
 remis
 au
 Service
 de
 la
 gestion
 des
 documents
 et
 des
 archives
 de
 l’Université
 et
 l’autre
 sera
 conservé
 au
 Département d'études cinématographiques.

Dialogue sur l'avenir de la démocratie libérale : John Rawls, Richard Rorty et Leo Strauss

Réalisé en cotutelle avec l'Université de Paris-Sorbonne (Paris IV).

Advances in scaling deep learning algorithms

Les algorithmes d'apprentissage profond forment un nouvel ensemble de méthodes puissantes pour l'apprentissage automatique. L'idée est de combiner des couches de facteurs latents en hierarchies. Cela requiert souvent un coût computationel plus elevé et augmente aussi le nombre de paramètres du modèle. Ainsi, l'utilisation de ces méthodes sur des problèmes à plus grande échelle demande de réduire leur coût et aussi d'améliorer leur régularisation et leur optimization. Cette thèse adresse cette question sur ces trois perspectives. Nous étudions tout d'abord le problème de réduire le coût de certains algorithmes profonds. Nous proposons deux méthodes pour entrainer des machines de Boltzmann restreintes et des auto-encodeurs débruitants sur des distributions sparses à haute dimension. Ceci est important pour l'application de ces algorithmes pour le traitement de langues naturelles. Ces deux méthodes (Dauphin et al., 2011; Dauphin and Bengio, 2013) utilisent l'échantillonage par importance pour échantilloner l'objectif de ces modèles. Nous observons que cela réduit significativement le temps d'entrainement. L'accéleration atteint 2 ordres de magnitude sur plusieurs bancs d'essai. Deuxièmement, nous introduisont un puissant régularisateur pour les méthodes profondes. Les résultats expérimentaux démontrent qu'un bon régularisateur est crucial pour obtenir de bonnes performances avec des gros réseaux (Hinton et al., 2012). Dans Rifai et al. (2011), nous proposons un nouveau régularisateur qui combine l'apprentissage non-supervisé et la propagation de tangente (Simard et al., 1992). Cette méthode exploite des principes géometriques et permit au moment de la publication d'atteindre des résultats à l'état de l'art. Finalement, nous considérons le problème d'optimiser des surfaces non-convexes à haute dimensionalité comme celle des réseaux de neurones. Tradionellement, l'abondance de minimum locaux était considéré comme la principale difficulté dans ces problèmes. Dans Dauphin et al. (2014a) nous argumentons à partir de résultats en statistique physique, de la théorie des matrices aléatoires, de la théorie des réseaux de neurones et à partir de résultats expérimentaux qu'une difficulté plus profonde provient de la prolifération de points-selle. Dans ce papier nous proposons aussi une nouvelle méthode pour l'optimisation non-convexe.

Chemical sensing with microbent optical fiber

We propose and demonstrate the possibility of using a permanently microbent bare optical fiber for detecting chemical species. Two detection schemes, viz., a bright-field detection scheme (for the core modes), and a dark-field detection scheme (for the cladding modes) have been employed to produce a fiber-optic sensor. The sensor described here is sensitive enough to detect concentrations as low as nanomoles per liter of a chemical species, with a dynamic range of more than 6 orders of magnitude.

Long period gratings in multimode optical fibers: application in chemical sensing

We propose and demonstrate a new technique for evanescent wave chemical sensing by writing long period gratings in a bare multimode plastic clad silica fiber. The sensing length of the present sensor is only 10 mm, but is as sensitive as a conventional unclad evanescent wave sensor having about 100 mm sensing length. The minimum measurable concentration of the sensor reported here is 10 nmol/l and the operating range is more than 4 orders of magnitude. Moreover, the detection is carried out in two independent detection configurations viz., bright field detection scheme that detects the core-mode power and dark field detection scheme that detects the cladding mode power. The use of such a double detection scheme definitely enhances the reliability and accuracy of the results. Furthermore, the cladding of the present fiber need not be removed as done in conventional evanescent wave fiber sensors.

Edge excitations and topological order in rotating Bose gas

The edge excitations and related topological orders of correlated states of a fast rotating Bose gas are studied. Using exact diagonalization of small systems, we compute the energies and number of edge excitations, as well as the boson occupancy near the edge for various states. The chiral Luttinger-liquid theory of Wen is found to be a good description of the edges of the bosonic Laughlin and other states identified as members of the principal Jain sequence for bosons. However, we find that in a harmonic trap the edge of the state identified as the Moore-Read (Pfaffian) state shows a number of anomalies. An experimental way of detecting these correlated states is also discussed.

Polyconductivity in polypyrrole: The correlated electron gas

p-toluensulfonate doped polypyrrole (PPy), undergoes an electric-field induced reversible transition from an insulating state to a highly conductive one. The spatially average field can be as small as 200 V/cm, when the temperature of the sample is below 20 K. The applied electric field leads to a sharp jump in the value of the current to a value which is nearly five orders of magnitude higher than before. When the applied electric field is reduced to below a critical value, the system switches back to a low conductive state. The effect is reversible, symmetric in voltage, and reproducible for different samples. The switching is, we believe, an electronic glass melting transition and it is due to the disordered, highly charged granular nature of PPy.

Accurate Hartree-Fock-Slater calculations on small diatomic molecules with the finite-element method

We report on the self-consistent field solution of the Hartree-Fock-Slater equations using the finite-element method for the three small diatomic molecules N_2, BH and CO as examples. The quality of the results is not only better by two orders of magnitude than the fully numerical finite difference method of Laaksonen et al. but the method also requires a smaller number of grid points.

Synthese und Charakterisierung neuer symmetrischer und unsymmetrischer Spiro-p-oligophenyle

In der vorliegenden Arbeit wurden neue symmetrische Spiro-p-oligophenyle der allgemeinen Form Spiro-o-Φ[n,n] mit der Gesamtkettenlänge o=2n+2 Phenylringen (o > 10) und der Zahl n der Phenylringe in den p-Oligophenylsubstituenten am Spirobifluorenkern, dargestellt. Neben den symmetrischen Verbindungen wurden erstmals auch unsymmetrische Spiro-p-oligophenyle der allgemeinen Form Spiro-o-Φ[n,m] mit o=n+m+2 (o = 3-7) und n ≠ m synthetisiert. Aufgrund der sehr geringen Löslichkeit der größeren Verbindungen wurden löslichkeitssteigernde Substituenten an den endständigen Phenylringen angebracht. Bei den Verbindungen, die mit Trimethylsilyl-Gruppen (TMS-) in den endständigen meta-Positionen „3“ und „5“ substituiert wurden, konnte die Löslichkeit um mehrere Größenordnungen gesteigert werden, sodass die Darstellung der symmetrischen Verbindungen bis zu einer Kettenlänge von 16 Phenylringen möglich wurde. Nach erfolgreicher Synthese und Aufreinigung wurden die TMS-Gruppen wieder entfernt und die erhaltenen, unsubstituierten Verbindungen charakterisiert. Zusätzlich wurden auch die TMS-Derivate untersucht. Zur Charakterisierung zählten neben der Reinheits- und Strukturanalytik unter anderem auch spektroskopische (UV/Vis-Absorption, Fluoreszenz, Fluoreszenzquantenausbeute), elektrochemische (Cyclovoltammetrie) und thermische (Thermogravimetrie, Dynamische Differenzkalorimetrie) Untersuchungen. Hier wurde unter anderem der Einfluss der Kettenlänge und der Position der Spiroverknüpfung auf isomere Verbindungen gleicher Kettenlänge untersucht. Bei den spektroskopischen Messungen konnte eine Konvergenz der längstwelligen Absorptionsbanden, bzw. kürzestwelligen Fluoreszenzbanden mit zunehmender Kettenlänge beobachtet werden. Die effektive Konjugationslänge konnte so aus experimentellen Daten bestimmt werden zu 12 Phenylringen in der Absorption und 14 Phenylringen in der Fluoreszenz. Bei den Isomeren gleicher Kettenlänge zeigte sich in der Absorption eine hypsochrome Verschiebung der Absorptionsmaxima mit zunehmender Verschiebung der Spiroverknüpfung zum Kettenende hin, während die Position der Spiroverknüpfung keinen messbaren Einfluss auf die Verschiebung der Fluoreszenzbanden hatte. Die Substitution mit TMS in den meta-Positionen zeigte keinen messbaren Einfluss auf die Absorptions- bzw. Fluoreszenzbanden. Die elektrochemischen Untersuchungen zeigten mit zunehmender Kettenlänge eine erleichterte Oxidation und Reduktion, während bei Isomeren gleicher Kettenlänge die Oxidation mit Verschiebung der Spiroverknüpfung zum Kettenende hin erschwert und die Reduktion erleichtert war. Die thermogravimetrischen Analysen (TGA) zeigten eine außerordentlich hohe thermische Stabilität (5% Massenabnahme unter Schutzgas) der Spiro-p-oligophenyle von Td,5% = 474°C bei Spiro-5Φ[1,2] bis 570°C bei Spiro 8Φ[3,3]. Ebenso blieben hohe Rückstandsmassen unter Schutzgas bei 850°C zurück, wie das Beispiel Spiro 8Φ[3,3] mit 68% zeigt. Die Verbindungen zeigten hohe Schmelzpunkte (max. 496°C bei Spiro-6Φ[0,4]) und Glasübergangstemperaturen (max. 434°C bei p-TMS-Spiro-8Φ[3,3]). Viele der Verbindungen, besonders die in den meta-Positionen TMS-substituierten Verbindungen, bildeten stabile amorphe Gläser.

Parameter Estimation in Chaotic Systems

This report examines how to estimate the parameters of a chaotic system given noisy observations of the state behavior of the system. Investigating parameter estimation for chaotic systems is interesting because of possible applications for high-precision measurement and for use in other signal processing, communication, and control applications involving chaotic systems. In this report, we examine theoretical issues regarding parameter estimation in chaotic systems and develop an efficient algorithm to perform parameter estimation. We discover two properties that are helpful for performing parameter estimation on non-structurally stable systems. First, it turns out that most data in a time series of state observations contribute very little information about the underlying parameters of a system, while a few sections of data may be extraordinarily sensitive to parameter changes. Second, for one-parameter families of systems, we demonstrate that there is often a preferred direction in parameter space governing how easily trajectories of one system can "shadow'" trajectories of nearby systems. This asymmetry of shadowing behavior in parameter space is proved for certain families of maps of the interval. Numerical evidence indicates that similar results may be true for a wide variety of other systems. Using the two properties cited above, we devise an algorithm for performing parameter estimation. Standard parameter estimation techniques such as the extended Kalman filter perform poorly on chaotic systems because of divergence problems. The proposed algorithm achieves accuracies several orders of magnitude better than the Kalman filter and has good convergence properties for large data sets.

Parallel Coupled Micro-Macro Actuators

This thesis presents a new actuator system consisting of a micro-actuator and a macro-actuator coupled in parallel via a compliant transmission. The system is called the Parallel Coupled Micro-Macro Actuator, or PaCMMA. In this system, the micro-actuator is capable of high bandwidth force control due to its low mass and direct-drive connection to the output shaft. The compliant transmission of the macro-actuator reduces the impedance (stiffness) at the output shaft and increases the dynamic range of force. Performance improvement over single actuator systems was expected in force control, impedance control, force distortion and reduction of transient impact forces. A set of quantitative measures is proposed and the actuator system is evaluated against them: Force Control Bandwidth, Position Bandwidth, Dynamic Range, Impact Force, Impedance ("Backdriveability'"), Force Distortion and Force Performance Space. Several theoretical performance limits are derived from the saturation limits of the system. A control law is proposed and control system performance is compared to the theoretical limits. A prototype testbed was built using permanenent magnet motors and an experimental comparison was performed between this actuator concept and two single actuator systems. The following performance was observed: Force bandwidth of 56Hz, Torque Dynamic Range of 800:1, Peak Torque of 1040mNm, Minimum Torque of 1.3mNm. Peak Impact Force was reduced by an order of magnitude. Distortion at small amplitudes was reduced substantially. Backdriven impedance was reduced by 2-3 orders of magnitude. This actuator system shows promise for manipulator design as well as psychophysical tests of human performance.

Reconfigurable Architectures for General-Purpose Computing

General-purpose computing devices allow us to (1) customize computation after fabrication and (2) conserve area by reusing expensive active circuitry for different functions in time. We define RP-space, a restricted domain of the general-purpose architectural space focussed on reconfigurable computing architectures. Two dominant features differentiate reconfigurable from special-purpose architectures and account for most of the area overhead associated with RP devices: (1) instructions which tell the device how to behave, and (2) flexible interconnect which supports task dependent dataflow between operations. We can characterize RP-space by the allocation and structure of these resources and compare the efficiencies of architectural points across broad application characteristics. Conventional FPGAs fall at one extreme end of this space and their efficiency ranges over two orders of magnitude across the space of application characteristics. Understanding RP-space and its consequences allows us to pick the best architecture for a task and to search for more robust design points in the space. Our DPGA, a fine- grained computing device which adds small, on-chip instruction memories to FPGAs is one such design point. For typical logic applications and finite- state machines, a DPGA can implement tasks in one-third the area of a traditional FPGA. TSFPGA, a variant of the DPGA which focuses on heavily time-switched interconnect, achieves circuit densities close to the DPGA, while reducing typical physical mapping times from hours to seconds. Rigid, fabrication-time organization of instruction resources significantly narrows the range of efficiency for conventional architectures. To avoid this performance brittleness, we developed MATRIX, the first architecture to defer the binding of instruction resources until run-time, allowing the application to organize resources according to its needs. Our focus MATRIX design point is based on an array of 8-bit ALU and register-file building blocks interconnected via a byte-wide network. With today's silicon, a single chip MATRIX array can deliver over 10 Gop/s (8-bit ops). On sample image processing tasks, we show that MATRIX yields 10-20x the computational density of conventional processors. Understanding the cost structure of RP-space helps us identify these intermediate architectural points and may provide useful insight more broadly in guiding our continual search for robust and efficient general-purpose computing structures.

Fast Object Recognition in Noisy Images Using Simulated Annealing

A fast simulated annealing algorithm is developed for automatic object recognition. The normalized correlation coefficient is used as a measure of the match between a hypothesized object and an image. Templates are generated on-line during the search by transforming model images. Simulated annealing reduces the search time by orders of magnitude with respect to an exhaustive search. The algorithm is applied to the problem of how landmarks, for example, traffic signs, can be recognized by an autonomous vehicle or a navigating robot. The algorithm works well in noisy, real-world images of complicated scenes for model images with high information content.