Deep learning of representations and its application to computer vision

L’objectif de cette thèse par articles est de présenter modestement quelques étapes du parcours qui mènera (on espère) à une solution générale du problème de l’intelligence artificielle. Cette thèse contient quatre articles qui présentent chacun une différente nouvelle méthode d’inférence perceptive en utilisant l’apprentissage machine et, plus particulièrement, les réseaux neuronaux profonds. Chacun de ces documents met en évidence l’utilité de sa méthode proposée dans le cadre d’une tâche de vision par ordinateur. Ces méthodes sont applicables dans un contexte plus général, et dans certains cas elles on tété appliquées ailleurs, mais ceci ne sera pas abordé dans le contexte de cette de thèse. Dans le premier article, nous présentons deux nouveaux algorithmes d’inférence variationelle pour le modèle génératif d’images appelé codage parcimonieux “spike- and-slab” (CPSS). Ces méthodes d’inférence plus rapides nous permettent d’utiliser des modèles CPSS de tailles beaucoup plus grandes qu’auparavant. Nous démontrons qu’elles sont meilleures pour extraire des détecteur de caractéristiques quand très peu d’exemples étiquetés sont disponibles pour l’entraînement. Partant d’un modèle CPSS, nous construisons ensuite une architecture profonde, la machine de Boltzmann profonde partiellement dirigée (MBP-PD). Ce modèle a été conçu de manière à simplifier d’entraînement des machines de Boltzmann profondes qui nécessitent normalement une phase de pré-entraînement glouton pour chaque couche. Ce problème est réglé dans une certaine mesure, mais le coût d’inférence dans le nouveau modèle est relativement trop élevé pour permettre de l’utiliser de manière pratique. Dans le deuxième article, nous revenons au problème d’entraînement joint de machines de Boltzmann profondes. Cette fois, au lieu de changer de famille de modèles, nous introduisons un nouveau critère d’entraînement qui donne naissance aux machines de Boltzmann profondes à multiples prédictions (MBP-MP). Les MBP-MP sont entraînables en une seule étape et ont un meilleur taux de succès en classification que les MBP classiques. Elles s’entraînent aussi avec des méthodes variationelles standard au lieu de nécessiter un classificateur discriminant pour obtenir un bon taux de succès en classification. Par contre, un des inconvénients de tels modèles est leur incapacité de générer deséchantillons, mais ceci n’est pas trop grave puisque la performance de classification des machines de Boltzmann profondes n’est plus une priorité étant donné les dernières avancées en apprentissage supervisé. Malgré cela, les MBP-MP demeurent intéressantes parce qu’elles sont capable d’accomplir certaines tâches que des modèles purement supervisés ne peuvent pas faire, telles que celle de classifier des données incomplètes ou encore celle de combler intelligemment l’information manquante dans ces données incomplètes. Le travail présenté dans cette thèse s’est déroulé au milieu d’une période de transformations importantes du domaine de l’apprentissage à réseaux neuronaux profonds qui a été déclenchée par la découverte de l’algorithme de “dropout” par Geoffrey Hinton. Dropout rend possible un entraînement purement supervisé d’architectures de propagation unidirectionnel sans être exposé au danger de sur- entraînement. Le troisième article présenté dans cette thèse introduit une nouvelle fonction d’activation spécialement con ̧cue pour aller avec l’algorithme de Dropout. Cette fonction d’activation, appelée maxout, permet l’utilisation de aggrégation multi-canal dans un contexte d’apprentissage purement supervisé. Nous démontrons comment plusieurs tâches de reconnaissance d’objets sont mieux accomplies par l’utilisation de maxout. Pour terminer, sont présentons un vrai cas d’utilisation dans l’industrie pour la transcription d’adresses de maisons à plusieurs chiffres. En combinant maxout avec une nouvelle sorte de couche de sortie pour des réseaux neuronaux de convolution, nous démontrons qu’il est possible d’atteindre un taux de succès comparable à celui des humains sur un ensemble de données coriace constitué de photos prises par les voitures de Google. Ce système a été déployé avec succès chez Google pour lire environ cent million d’adresses de maisons.

Sub-Lethal effects of heavy metals on perna indica ( Kuriakose & Nair) and donax incarnatus gmelin

The present scientific investigation of the effects of copper, mercury and cadmium has focussed on their effects on two commercially important marine bivalve species, Perna indica (brown mussel) and Donax incarnatus (wedge clam), conspicuous representatives of the tropical intertidal areas. The investigation centred around delineating the cause and effects of heavy metal stress, individually and in combination on these species under laboratory conditions. A clear understanding of the cause and effect can be had only if laboratory experiments are conducted employing sub-lethal concentrations of the above toxicants. Therefore, during the course of the investigation, sub-lethal concentrations of copper, mercury and cadmium were employed to assess the concentration dependent effects on survival, ventilation rate, O:N ratio and tissues. The results obtained are compared with the already available information and partitioned in sections to make a meaningful presentation.The thesis is presented in five chapters comprising INTRODUCTION, ACUTE TOXICITY, VENTILATION RATE, OXYGEN : NITROGEN RATIO and HISTOPATHOLOGY. Each chapter has been divided into various sections such as INTRODUCTION, REVIEW OF LITERATURE, MATERIAL AND METHODS, RESULTS and DISCUSSION

A Comparative Study Of Tensile And Microstructural Characteristics Of Friction Welded Joints And Conventionally Welded Joints Of Aluminium Alloy 6061

Welding of high strength and low weight materials like Aluminium Alloys without any defects by conventional welding techniques is a major challenge in industries. Hence research on solid state welding techniques like Friction stir welding and Friction welding techniques have got much importance in joining of Aluminium alloys. However most of the industries are not changing conventional techniques as skilled workers are available on that area. Most common conventional welding techniques used for joining of Aluminium alloys are Gas welding and Arc welding. Friction welding is a solid-state welding process that generates heat through mechanical friction between a moving and a stationary component with the addition of a lateral force called “upset” to plast ically displace and fuse the materials. In this work, experimental study on tensile and micro structural characteristics of welded joints formed from conventional welding techniques and Rotary friction welding(suitable for weld specimens with circular cross section) has been carried out and the same were compared. The process parameters for arc welding used was 50-70 Amp reverse polarity DC and electrodes of 2.3mm diameter. In Gas welding, the parameters were oxy acetylene neural flame at 3200°C and 3mm electrodes . In the case of friction welding an axial pressure loading of 3Mpa with 5 MPa as upsetting pressure and 500 rpm were used to obtain good welded joints. Tensile characteristic studies of Arc welded joints and Gas welded joints showed 48% and 60 % variations respectively from the maximum load bearing characteristics of parent metal. In the case of friction welded joint, the variation was found to 46%. Micro structural evaluation of conventionally welded joints exhibited clear distinct zones of various weld regions. In the case of friction welded joint micro structural photographs showed comparable features both in parent metal and welded region. Thus the tensile characteristic study and microstructure evaluations proved that friction welded joints are good in both aspects compared to conventionally welded joints.

Gender and Space in Tehran

Abstract: This dissertation generally concentrates on the relationships between “gender” and “space” in the present time of urban life in capital city of Tehran. “Gender” as a changing social construct, differentiated within societies and through time, studied this time by investigation on “gender attitude” or “gender identity” means attitudes towards “gender” issues regarding Tehran residences. “Space” as a concept integrated from physical and social constituents investigated through focus on “spatial attitude” means attitudes towards using “living spaces” including private space of “house”, semi private semi public space of neighborhood and finally public spaces of the city. “Activities and practices” in space concentrated instead of “physical” space; this perspective to “space” discussed as the most justified implication of “space” in this debate regarding current situations in city of Tehran. Under a systematic approach, the interactions and interconnections between “gender” and “space” as two constituent variables of social organization investigated by focus on the different associations presented between different “gender identities” and their different “spatial identities”; in fact, “spatial identity” manifests “gender identity” and in opposite direction, “spatial identity” influences to construction of “gender identity”. The hypotheses of case study in Tehran defined as followed: • “Gender identity” is reflected on “spatial identity”. Various “gender identities” in Tehran present different perspectives of “space” or they identify “space” by different values. • As “gender identity” internalizes patriarchal oppression, it internalizes associated “spatial” oppression too. • Within the same social class, different “gender identities” related to men and women, present interconnected qualities, compared with “gender identities” related to men or women of different social classes. This situation could be found in the “spatial” perspectives of different groups of men and women too. • Following the upper hypotheses, “spatial” oppression differs among social classes of Tehran living in different parts of this city. This research undertook a qualitative study in Tehran by interviewing with different parents of both young daughter and son regarding their attitudes towards gender issues from one side and activities and behaviors of their children in different spaces from the other side. Results of case study indicated the parallel changes of parents’ attitudes towards “gender” and “spatial” issues; it means strong connection between “gender” and “space”. It revealed association of “equal” spatial attitudes with “open, neutral” gender attitudes, and also the association of “biased, unequal” spatial identities with “conservative patriarchal” gender identities. It was cleared too that this variable concept – gender space - changes by “sex”; mothers comparing fathers presented more equitable notions towards “gender spatial” issues. It changes too by “social class” and “educational level”, that means “gender spatial” identity getting more open equitable among more educated people of middle and upper classes. “Breadwinning status in the family” also presents its effect on the changes of “gender spatial” identity so participant breadwinners in the family expressed relatively more equitable notions comparing householders and housekeepers. And finally, “gender spatial” identity changes through “place” in the city and regarding South – North line of the city. The illustration of changes of “gender spatial” identity from “open” to “conservative” among society indicated not only vertical variation across social classes, furthermore the horizontal changing among each social class. These results also confirmed hypotheses while made precision on the third one regarding variable of sex. More investigations pointed to some inclusive spatial attitudes throughout society penetrated to different groups of “gender identities”, to “opens” as to “conservatives”, also to groups between them, by two opposite features; first kind, conservative biased spatial practices in favor of patriarchal gender relations and the second, progressive neutral actions in favor of equal gender relations. While the major reason for the inclusive conservative practices was referred to the social insecurity for women, the second neutral ones associated to more formal & safer spaces of the city. In conclusion, while both trends are associated deeply with the important issues of “sex” & “body” in patriarchal thoughts, still strong, they are the consequences of the transitional period of social change in macro level, and the challenges involved regarding interactions between social orders, between old system of patriarchy, the traditional biased “gender spatial” relations and the new one of equal relations. The case study drew an inhomogeneous illustration regarding gender spatial aspects of life in Tehran, the opposite groups of “open” and “conservative”, and the large group of “semi open semi conservative” between them. In macro perspective it presents contradicted social groups according their general life styles; they are the manifestations of challenging trends towards tradition and modernity in Iranian society. This illustration while presents unstable social situations, necessitates probing solutions for social integration; exploring the directions could make heterogeneous social groups close in the way they think and the form they live in spaces. Democratic approaches like participatory development planning might be helpful for the city in its way to more solidarity and sustainability regarding its social spatial – gender as well – development, in macro levels of social spatial planning and in micro levels of physical planning, in private space of house and in public spaces of the city.

Extraversion and neuroticism as predictors of salivary cortisol levels in public speaking. 'Extraversión y neuroticismo como predictores de los niveles de cortisol salivar al hablar en público'.

Resumen basado en la publicación

Comparison of speech processor MAP minimum (T) and maximum (C) levels with psychophysical thresholds and maximum acceptable loudness levels in users of the Nucleus 22 cochlear implant system

This paper discusses the Nucleus 22 cochlear implant.

Physical and perceived noise levels within the neonatal intensive care unit of St. Louis Children's Hospital

This paper studies the success of the renovation of the Neonatal Intensive Care Unit at St. Louis Children's Hospital in reducing noise levels and improving the work environment.

Most comfortable listening levels of hearing impaired people: effects of everyday auditory and visual environments

This paper discusses a study to evaluate the changes that may occur in most comfortable listening levels when hearing aid users encounter a realistic auditory and visual environment.

Loudness discomfort levels: A retrospective study comparing data from Pascoe (1988) and Washington University School of Medicine

Loudess discomfort levels (LDLs) were gathered from three Washington University School of Medicine sites, for a total of 325 subjects (total ears=454). These levels were compared to mean LDLs reported by Pascoe (1988). The results revealed that the mean LDL measured at WUSM (ie., the IHAFF procedure) is significantly different that the LDL reported by Pascoe (1988).

Dietary supplements of whole linseed and vitamin E to increase levels of alpha-linolenic acid and vitamin E in bovine milk

The potential to increase the concentrations of n-3 polyunsaturated fatty acids (PUFAs) in milk fat was investigated by studying the effects of feeding a xylose-treated, whole cracked linseed supplement ( rich in alpha-linolenic acid) to dairy cows. Also the effect of increasing the dietary intake of vitamin E on the vitamin E status of milk was investigated. The effect of pasteurisation on milk fatty acid composition was also examined. Using a 3 x 2 factorial design, a total of 60 Holstein dairy cows were fed a total mixed ration based on grass silage supplemented with one of three levels of whole cracked linseed (78, 142 or 209 g . kg(-1) diet dry matter (DM); designated LL, ML or HL, respectively) in combination with one of two levels of additional dietary vitamin E intake ( 6 or 12 g vitamin E . animal(-1) . day(-1); designated LE or HE, respectively). Increasing lipid supplementation reduced (P < 0.01) diet DM intake and milk yield, and increased (P < 0.001) the overall content of oleic, vaccenic, alpha-linolenic and conjugated linoleic acids, and total PUFAs and monounsaturated fatty acids (MUFA). Myristic and palmitic acids in milk fat were reduced ( P < 0.001) through increased lipid supplementation. While α-linolenic acid concentrations were substantially increased this acid only accounted for 0.02 of total fatty acids in milk at the highest level of supplementation (630 g α-linolenic acid &BULL; animal(-1) &BULL; day(-1) for HL). Conjugated linoleic acid concentrations in milk fat were almost doubled by increasing the level of lipid supplementation (8.9, 10.4 and 16.1 g &BULL; kg(-1) fatty acids for LL, ML and HL, respectively). Although milk vitamin E contents were generally increased there was no benefit (P > 0.05) of increasing vitamin E intake from 6 to 12 g . animal(-1) . day(-1). The fatty acid composition of milk was generally not affected by pasteurisation.

Critical stages of the brewing process for changes in antioxidant activity and levels of phenolic compounds in ale

Samples were taken at each stage of brewing (malt, milling, mashing, wort separation, hop addition, boiling, whirlpool, dilution, fermentation, warm rest, chill-lagering, beer filtration, carbonation and bottling, pasteurization, and storage). The level of antioxidant activity of unfractionated, low-molecular-mass (LMM) and high-molecular-mass (HMM) fractions was measured by the 2,2'-azinobis(3-ethylbenzothiazoline-6-sulfortic acid) radical cation (ABTS(.+)) and ferric-reducing antioxidant power (FRAP) procedures. Polyphenol levels were assessed by HPLC. The LMM fraction (<5 kDa) was responsible for similar to80% of the level of antioxidant activity of the unfractionated malt and beer samples. In the unfractionated samples, significant decreases (P < 0.001) in antioxidant activity levels were observed after milling and beer filtration, with the decrease after beer filtration being accompanied by a significant decrease (P > 0.001) in catechin and ferulic acid levels. Increases in antioxidant activity levels were observed after mashing, boiling, fermentation, chill-lagering, and pasteurization, in line with previous studies on lager. Additionally, increases in the level of antioxidant activity occurred after wort separation and carbonation and bottling and were accompanied by increases in levels of most monitored polyphenols. Data from the ABTS(.-) and FRAP assays indicated that the compounds contributing to the levels of antioxidant activity responded differently in the two procedures. Levels of ferulic, vanillic, and chlorogenic acids and catechin accounted for 45-61% of the variation in antioxidant activity levels.

Effect of apoE genotype and dietary quercetin on blood lipids and TNF-alpha levels in apoE3 and apoE4 targeted gene replacement mice

The aim of this study was to determine the effect of dietary quercetin supplementation on blood lipids and TNF-alpha levels according to the apoE genotype in apoE3 and apoE4 targeted gene replacement mice. In a two-factorial design female apoE3 and apoE4 mice were fed semi-synthetic diets without (controls) and with quercetin (2 mg/g diet) for 6 weeks. Feeding the quercetin-supplemented diets significantly increased plasma levels of quercetin and isorhamnetin both in apoE3 and apoE4 mice. There was no significant effect of apoE genotype on plasma quercetin levels. ApoE3 and apoE4 transgenic mice exhibited similar plasma levels of apoE and cholesterol which were not significantly affected by dietary quercetin supplementation. In mice receiving the basal diet without quercetin supplementation, levels of TNF-alpha in whole blood stimulated ex vivo with lipopolysaccharide were higher in apoE3 as compared to apoE4 transgenic mice. Dietary quercetin significantly lowered levels of TNF-alpha by 44% in apoE3 mice relative to apoE3 mice receiving the unsupplemented diets. In apoE4 mice a moderate (20%) but not significant decrease in TNF-alpha levels in response to the quercetin supplementation was evident. Following quercetin supplementation TNF-alpha levels were similar between apoE3 and apoE4 transgenic mice. Current findings indicate that apoE3 mice are more responsive to the TNF-alpha lowering properties of dietary quercetin supplementation as compared to apoE4 animals.

Mothers' attributions following their child's diagnosis of autistic spectrum disorder: exploring links with maternal levels of stress, depression and expectations about their child's future

Although the impact of autism spectrum disorders (ASDs) on the family is well recognized, the way mothers attempt to make sense of the diagnosis is largely unexplored. However, in other disabilities, attributions have been shown to predict a variety of outcomes including maternal wellbeing and engagement in treatment. Using Weiner's (198S) three-dimensional model, 16 mothers were interviewed to examine the nature and impact of their beliefs about their child's ASD using semi-structured interviews and measures of depression, parenting stress and expectations for their child's future. The findings suggested that mothers made a diverse and complex range of attributions that were consistent with Weiner's dimensions of locus of cause, stability and controllability. The nature of their attributions reflected particular difficulties associated with ASDs, such as uncertainties regarding cause and prognosis. Taking account of mothers' search for meaning will better enable professionals to support families following diagnosis.