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.

A Mediated Presence : Uncovering Nietzsche in Art History from its Foundations to Contemporary Anglo-American Curricula

La pensée de Nietzsche a joué un rôle déterminant et récurrent dans les discours et les débats qui ont formé et continuent de façonner le domaine de l’histoire de l’art, mais aucune analyse systématique de cette question n’a encore vu le jour. L’influence de Nietzsche a été médiée par divers interlocuteurs, historiens de l’art et philosophes, qui ont encadré ces discussions, en utilisant les écrits du philosophe comme toile de fond de leurs propres idées. Ce mémoire souhaite démontrer que l’impact de Nietzsche dans le champ de l’histoire de l’art existe mais qu’il fut toujours immergé ou éclipsé, particulièrement dans le contexte anglo-américain, l’emphase étant placée sur les médiateurs de ses idées en n’avouant que très peu d’engagement direct avec son œuvre. En conséquence, son importance généalogique pour certains fondateurs de la discipline reste méconnue; sa présence réellement féconde se traduit plutôt comme une absence ou une présence masquée. En vue de démontrer ce propos, nous regardons donc le contexte nietzschéen qui travaille les écrits de certains historiens de l’art, comme Jacob Burckhardt et Aby Warburg, ou des philosophes et d’écrivains ayant marqué la discipline de l’histoire de l’art (plus particulièrement dans le cadre de l’influence de la « French Theory » sur l’histoire de l’art anglo-américaine depuis la fin des années 1970) : Martin Heidegger, Michel Foucault, Jacques Derrida, Gilles Deleuze et Georges Bataille. Nous examinons certaines voies par lesquelles ses idées ont acquis une pertinence pour l’histoire de l’art avant de proposer les raisons potentielles de leur occlusion ultérieure. Nous étudions donc l’évolution des discours multiples de l’histoire comme domaine d’étude afin de situer la contribution du philosophe et de cerner où et comment ses réflexions ont croisé celles des historiens de l’art qui ont soit élargi ou redéfini les méthodes et les structures d’analyse de leur discipline. Ensuite nous regardons « l’art » de Nietzsche en le comparant avec « l’art de l’histoire de l’art » (Preziosi 2009) afin d’évaluer si ces deux expressions peuvent se rejoindre ou s’il y a fondamentalement une incompatibilité entre les deux, laquelle pourrait justifier ou éclairer la distance entre la pensée nietzschéenne sur l’art et la discipline de l’histoire de l’art telle qu’elle s’institutionnalise au moment où le philosophe rédige son œuvre.

Is Breast Asymmetry Present in Girls with Adolescent Idiopathic Scoliosis?

Study Design Cross-sectional descriptive study. Objectives To characterize breast asymmetry (BA), as defined by breast volume difference, in girls with significant adolescent idiopathic scoliosis (AIS), using magnetic resonance imaging (MRI). Summary and Background BA is a frequent concern among girls with AIS. It is commonly believed that this results from chest wall deformity. Although many women exhibit physiological BA, the prevalence is not known in adolescents and it remains unclear if it is more frequent in AIS. Breasts vary in shape and size and many ways of measuring them have been explored. MRI shows the highest precision at defining breast tissue. Methods Thirty patients were enrolled on the basis of their thoracic curvature, skeletal and breast maturity, without regard to their perception on their BA. MRI acquisitions were performed in prone with a 1.5-Tesla system using a 16-channel breast coil. Segmentation was achieved using the ITK-SNAP 2.4.0 software and subsequently manually refined. Results The mean left breast volume (528.32 ± 205.96 cc) was greater compared with the mean right breast volume (495.18 ± 170.16 cc) with a significant difference between them. The mean BA was found to be 8.32% ± 6.43% (p < .0001). A weak positive correlation was observed between BA and thoracic Cobb angle (0.177, p = .349) as well as thoracic gibbosity angle (0.289, p = .122). The left breast was consistently larger in 65.5% of the patients. Twenty patients (66.7%) displayed BA ≥5%. Conclusions We have described BA in patients with significant AIS using MRI. This method is feasible, objective, and very precise. The majority of patients had a larger left breast, which could compound the apparent BA secondary to trunk rotation. In many cases, BA is present independently of thoracic deformity. This knowledge will assist in counseling AIS patients in regards to their concerns with BA.

Career Officers and Political Appointees in American Diplomacy: Contending for Diplomatic Positions through Boundary Work

American Political and Economic Series

Evanescent Wave Fibre Optic Sensors for Trace Analysis of Fe3+ in Water

In this communication, we discuss the details of fabricating an off-line fibre optic sensor (FOS) based on evanescent wave absorption for detecting trace amounts of Fe3+ in water. Two types of FOS are developed; one type uses the unclad portion of a multimode silica fibre as the sensing region whereas the other employs the microbent portion of a multimode plastic fibre as the sensing region. Sensing is performed by measuring the absorption of the evanescent wave in a reagent medium surrounding the sensing region. To evaluate the relative merits of the two types of FOS in Fe3+ sensing, a comparative study of the sensors is made, which reveals the superiority of the latter in many respects, such as smaller sensing length, use of a double detection scheme (for detecting both core and cladding modes) and higher sensitivity of cladding mode detection at an intermediate range of concentration along with the added advantage that plastic fibres are inexpensive. A detection limit of 1 ppb is observed in both types of fibre and the range of detection can be as large as 1 ppb–50 ppm. All the measurements are carried out using a LabVIEW set-up.

Evanescent wave fibre optic sensors for trace analysis of Fe3+ in water

In this communication, we discuss the details of fabricating an off-line fibre optic sensor (FOS) based on evanescent wave absorption for detecting trace amounts of Fe3+ in water. Two types of FOS are developed; one type uses the unclad portion of a multimode silica fibre as the sensing region whereas the other employs the microbent portion of a multimode plastic fibre as the sensing region. Sensing is performed by measuring the absorption of the evanescent wave in a reagent medium surrounding the sensing region. To evaluate the relative merits of the two types of FOS in Fe3+ sensing, a comparative study of the sensors is made, which reveals the superiority of the latter in many respects, such as smaller sensing length, use of a double detection scheme (for detecting both core and cladding modes) and higher sensitivity of cladding mode detection at an intermediate range of concentration along with the added advantage that plastic fibres are inexpensive. A detection limit of 1 ppb is observed in both types of fibre and the range of detection can be as large as 1 ppb–50 ppm. All the measurements are carried out using a LabVIEW set-up.

Evanescent wave fibre optic sensors for trace analysis of Fe3+ in water

In this communication, we discuss the details of fabricating an off-line fibre optic sensor (FOS) based on evanescent wave absorption for detecting trace amounts of Fe3+ in water. Two types of FOS are developed; one type uses the unclad portion of a multimode silica fibre as the sensing region whereas the other employs the microbent portion of a multimode plastic fibre as the sensing region. Sensing is performed by measuring the absorption of the evanescent wave in a reagent medium surrounding the sensing region. To evaluate the relative merits of the two types of FOS in Fe3+ sensing, a comparative study of the sensors is made, which reveals the superiority of the latter in many respects, such as smaller sensing length, use of a double detection scheme (for detecting both core and cladding modes) and higher sensitivity of cladding mode detection at an intermediate range of concentration along with the added advantage that plastic fibres are inexpensive. A detection limit of 1 ppb is observed in both types of fibre and the range of detection can be as large as 1 ppb–50 ppm. All the measurements are carried out using a LabVIEW set-up.

Evanescent wave fibre optic sensors for trace analysis of Fe3+ in water

In this communication, we discuss the details of fabricating an off-line fibre optic sensor (FOS) based on evanescent wave absorption for detecting trace amounts of Fe3+ in water. Two types of FOS are developed; one type uses the unclad portion of a multimode silica fibre as the sensing region whereas the other employs the microbent portion of a multimode plastic fibre as the sensing region. Sensing is performed by measuring the absorption of the evanescent wave in a reagent medium surrounding the sensing region. To evaluate the relative merits of the two types of FOS in Fe3+ sensing, a comparative study of the sensors is made, which reveals the superiority of the latter in many respects, such as smaller sensing length, use of a double detection scheme (for detecting both core and cladding modes) and higher sensitivity of cladding mode detection at an intermediate range of concentration along with the added advantage that plastic fibres are inexpensive. A detection limit of 1 ppb is observed in both types of fibre and the range of detection can be as large as 1 ppb–50 ppm. All the measurements are carried out using a LabVIEW set-up.

A Beacon based Localization System for Autonomous Mobile Robots and its Applications in Traffic and Transport Control

ACCURATE sensing of vehicle position and attitude is still a very challenging problem in many mobile robot applications. The mobile robot vehicle applications must have some means of estimating where they are and in which direction they are heading. Many existing indoor positioning systems are limited in workspace and robustness because they require clear lines-of-sight or do not provide absolute, driftfree measurements.The research work presented in this dissertation provides a new approach to position and attitude sensing system designed specifically to meet the challenges of operation in a realistic, cluttered indoor environment, such as that of an office building, hospital, industrial or warehouse. This is accomplished by an innovative assembly of infrared LED source that restricts the spreading of the light intensity distribution confined to a sheet of light and is encoded with localization and traffic information. This Digital Infrared Sheet of Light Beacon (DISLiB) developed for mobile robot is a high resolution absolute localization system which is simple, fast, accurate and robust, without much of computational burden or significant processing. Most of the available beacon's performance in corridors and narrow passages are not satisfactory, whereas the performance of DISLiB is very encouraging in such situations. This research overcomes most of the inherent limitations of existing systems.The work further examines the odometric localization errors caused by over count readings of an optical encoder based odometric system in a mobile robot due to wheel-slippage and terrain irregularities. A simple and efficient method is investigated and realized using an FPGA for reducing the errors. The detection and correction is based on redundant encoder measurements. The method suggested relies on the fact that the wheel slippage or terrain irregularities cause more count readings from the encoder than what corresponds to the actual distance travelled by the vehicle.The application of encoded Digital Infrared Sheet of Light Beacon (DISLiB) system can be extended to intelligent control of the public transportation system. The system is capable of receiving traffic status input through a GSM (Global System Mobile) modem. The vehicles have infrared receivers and processors capable of decoding the information, and generating the audio and video messages to assist the driver. The thesis further examines the usefulness of the technique to assist the movement of differently-able (blind) persons in indoor or outdoor premises of his residence.The work addressed in this thesis suggests a new way forward in the development of autonomous robotics and guidance systems. However, this work can be easily extended to many other challenging domains, as well.

Constraining the active tectonic deformation of Andaman-Nicobar Arc in the background of December 26,2004 Great Sumatra-Andaman Earthquake

The Andaman-Nicobar Islands in the Bay of Bengal lies in a zone where the Indian plate subducts beneath the Burmese microplate, and therefore forms a belt of frequent earthquakes. Few efforts, not withstanding the available historical and instrumental data were not effectively used before the Mw 9.3 Sumatra-Andaman earthquake to draw any inference on the spatial and temporal distribution of large subduction zone earthquakes in this region. An attempt to constrain the active crustal deformation of the Andaman-Nicobar arc in the background of the December 26, 2004 Great Sumatra-Andaman megathrust earthquake is made here, thereby presenting a unique data set representing the pre-seismic convergence and co-seismic displacement.Understanding the mechanisms of the subduction zone earthquakes is both challenging sCientifically and important for assessing the related earthquake hazards. In many subduction zones, thrust earthquakes may have characteristic patterns in space and time. However, the mechanism of mega events still remains largely unresolved.Large subduction zone earthquakes are usually associated with high amplitude co-seismic deformation above the plate boundary megathrust and the elastic relaxation of the fore-arc. These are expressed as vertical changes in land level with the up-dip part of the rupture surface uplifted and the areas above the down-dip edge subsided. One of the most characteristic pattern associated with the inter-seismic era is that the deformation is in an opposite sense that of co-seismic period.This work was started in 2002 to understand the tectonic deformation along the Andaman-Nicobar arc using seismological, geological and geodetic data. The occurrence of the 2004 megathrust earthquake gave a new dimension to this study, by providing an opportunity to examine the co-seismic deformation associated with the greatest earthquake to have occurred since the advent of Global Positioning System (GPS) and broadband seismometry. The major objectives of this study are to assess the pre-seismic stress regimes, to determine the pre-seismic convergence rate, to analyze and interpret the pattern of co-seismic displacement and slip on various segments and to look out for any possible recurrence interval for megathrust event occurrence for Andaman-Nicobar subduction zone. This thesis is arranged in six chapters with further subdivisions dealing all the above aspects.

Biochemical Studies on the Protective Effect of Taurine on Experimentally Induced Fulminat Hepatic Failure in Rats

Fulminant hepatic failure (FHF) is a dramatic and challenging syndrome in clinical medicine. Although an uncommon disorder, it is usually fatal and occurs in previously healthy person. While the causes of FHF remain unclear, viral hepatitis and drug-induced liver injury account for the majority of cases. Hepatitis E causes large-scale epidemics of hepatitis in the Indian subcontinent, involving hundreds of thousands of cases with high mortality. FHF is associated with several clinical features like jaundice, shrunken liver, easy bruising, low levels of serum proteins, fatigue, multi-organ failure etc and metabolic derangements like hypoglycemia, hyperlipidemia, hyponatremia, defective protein synthesis, reduced energy production, decreased rate of urea production etc. These disturbances are predominantly attributed to oxidative stress, membrane destabilization and osmolytic imbalances. The options available for these patients are quite minimal with liver transplantation being one of them. But the procedure is ridden with issues causing it to find less favor among the patients and the caregivers. Use of hepatoprotective and cytoprotective drugs, is being considered to be a more acceptable alternative as a strategy to enhance liver regeneration. In this regard use of taurine a naturally occurring amino acid that plays a crucial role in many physiological processes would prove to be effective. In the present study, hepatoprotective effect of taurine on a rat model of induced FHF was studied. Taurine supplementation has effectively counteracted the metabolic and structural aberrations in the liver caused by D-galactosamine intoxication.

2(3H)-Furanones and related substrates: Synthetic strategies and selectivity profile in their reactivity

The synthesis and reactions of simple derivatives of 2(3H)- and 3(2H)furanones have attracted considerable attention in recent years, primarily in connection with development of routes to antitumor agents that contain this ring as central structural unit. They also serve as useful synthetic building blocks for lactones and furans and are the precursors of a wide variety of biologically important heterocyclic systems. Although a number of syntheses of furanones were known they were in many cases limited to specific substitution pattems. The development of altemative strategies for the preparation of these heterocycles is therefore of considerable importance or continues to be a challenge.We propose to develop new and general approaches to the synthesis of furanone ring systems from simple and readily available starting materials since we were interested in examining their rich photochemistry. The photochemical reactivity of Beta,gama-unsaturated lactams and lactones is a subject of current interest. Some of the prominent photoreaction pathways of unsaturated lactones include decarbonylation, solvent addition to double bonds, decarboxylation, migration of aryl substituents and dimerisation. lt was reported earlier that the critical requirement for clean photochemical cleavage of the acyl-oxygen bond is the presence ofa double bond adjacent to the ether oxygen and 2(3H)-furanones possessing this structural requirement undergo facile decarbonylation. But related phenanthrofuranones are isolated as photostable end products upon irradiation. Hence we propose to synthesis a few phenanthro-2(3H)-furanones to study the effect of a radical stabilising group at 3-position of furanone ring on photolysis. To explore the tripletmediated transformations of 2(3H)-furanones in polar and nonpolar solvents a few 3,3-bis(4-chlorophenyl)-5-aryl-3H-furan-2-ones and 3,3-di(p-tolyl)-5-aryl- 3H-furan-2-ones were synthesised from the corresponding dibenzoylstyrene precursors by neat thermolysis. Our aim was to study the nature of intermediates involved in these transformations.We also explored the possibility of developing a new and general approach to the synthesis of 3(2H)-furanones from simple and readily available starting materials since such general procedures are not available. The protocol developed by us employs readily available phenanthrenequinone and various 4-substituted acetophenones as starting materials and provides easy access to the required 3(2H)-furanone targets. These furanone derivatives have immense potential for further investigations .We also aimed the synthesis of a few dibenzoylalkene-type systems such as acenaphthenone-2—ylidene ketones and phenanthrenone-9-ylidene ketones. These systems were expected to undergo thermal rearrangement to give furanones and spirofuranones. Also these systems can be categorised as quinonemethides which are valuable synthetic intermediates.

Ecophysiology of nontuberculous Mycobacteria in marine aquaculture ponds

The world demand for fish and fishery products is increasing steadily and it is generally accepted that it will not be possible to meet the heavy demand with resources exploited from capture fishery alone. Now aquaculture is well established and fastdeveloping industry in many countries and is a major focus sector for development. During recent decades, aquaculture has gained momentum, throughout the world especially in developing countries. According to Food and Agricultural Oganisation (FAO, 2000), global aquaculture production was 26.38 tones in 1996 have reached 32.9 million tonnes during 1999. Only marine aquaculture sector has contributed 13.1 million tonnes during 1999.India is a major fish producing country. About one half of lndia’s brackish water lands are currently being utilized for farming in order to reduce the gap between supply and demand for fish. Aquaculture has become a major source of livelihood for people and its role in integrated rural development, generation of employment and earning foreign exchange, thereby alleviating poverty is being greatly appreciated around the world.Among the infectious agents, bacteria are becoming the prime causal organisms for diseases in food fishes and other marine animals. Sindermann, (1970) reported that bacterial fish pathogen most commonly found among marine fishes is species of Pseudomonas, Vibrio and Mycobacterium. These can be categorized into primary pathogens; secondary invaders that may cause systemic disease in immunocompromised hosts; and normal marine flora which are not pathogenic but may occur on body surfaces or even within the tissues of the host. I-Iigh density of animals in hatchery tanks and ponds is conducive to the spread of pathogen and the aquatic environment with regular application of protein rich feed, is ideal for culturing bacteria. Bacteria, which are normally present in seawater or on the surface of fish, can invade and cause pathological effects in fishes, which are injured or subjected to other environmental stresses.Mycobacteria except parasites are known as nontuberculosis mycobacteria (NTM), atypical mycobacteria or mycobacteria other than tuberculosis(MO'l'l"). This group of mycobacteria includes opportunistic pathogens and saprophytes. Environmental mycobacteria are ubiquitous in distribution and the sources may include soil, water, warm-blooded as well as cold-blooded animals. Disease caused by environmental mycobacterial strains in susceptible humans (Goslee & Wolinsky, 1976; Grange, 1987), animals and fishes are increasingly attracting attention. Greatest importance of environmental mycobacteria is believed to be their role in immunological priming of humans and animals, thereby modifying their immune responses to subsequent exposure to pathogenic species.

Nonparametric Estimation of Survivor Function in Bivariate Competing Risk Models’

So far, in the bivariate set up, the analysis of lifetime (failure time) data with multiple causes of failure is done by treating each cause of failure separately. with failures from other causes considered as independent censoring. This approach is unrealistic in many situations. For example, in the analysis of mortality data on married couples one would be interested to compare the hazards for the same cause of death as well as to check whether death due to one cause is more important for the partners’ risk of death from other causes. In reliability analysis. one often has systems with more than one component and many systems. subsystems and components have more than one cause of failure. Design of high-reliability systems generally requires that the individual system components have extremely high reliability even after long periods of time. Knowledge of the failure behaviour of a component can lead to savings in its cost of production and maintenance and. in some cases, to the preservation of human life. For the purpose of improving reliability. it is necessary to identify the cause of failure down to the component level. By treating each cause of failure separately with failures from other causes considered as independent censoring, the analysis of lifetime data would be incomplete. Motivated by this. we introduce a new approach for the analysis of bivariate competing risk data using the bivariate vector hazard rate of Johnson and Kotz (1975).

Some Lattice Problems In Fuzzy Set Theory And Fuzzy Topology

It is believed that every fuzzy generalization should be formulated in such a way that it contain the ordinary set theoretic notion as a special case. Therefore the definition of fuzzy topology in the line of C.L.CHANG E9] with an arbitrary complete and distributive lattice as the membership set is taken. Almost all the results proved and presented in this thesis can, in a sense, be called generalizations of corresponding results in ordinary set theory and set topology. However the tools and the methods have to be in many of the cases, new. Here an attempt is made to solve the problem of complementation in the lattice of fuzzy topologies on a set. It is proved that in general, the lattice of fuzzy topologies is not complemented. Complements of some fuzzy topologies are found out. It is observed that (L,X) is not uniquely complemented. However, a complete analysis of the problem of complementation in the lattice of fuzzy topologies is yet to be found out