Yeast vacuoles fragment in an asymmetrical two-phase process with distinct protein requirements.

Yeast vacuoles fragment and fuse in response to environmental conditions, such as changes in osmotic conditions or nutrient availability. Here we analyze osmotically induced vacuole fragmentation by time-lapse microscopy. Small fragmentation products originate directly from the large central vacuole. This happens by asymmetrical scission rather than by consecutive equal divisions. Fragmentation occurs in two distinct phases. Initially, vacuoles shrink and generate deep invaginations that leave behind tubular structures in their vicinity. Already this invagination requires the dynamin-like GTPase Vps1p and the vacuolar proton gradient. Invaginations are stabilized by phosphatidylinositol 3-phosphate (PI(3)P) produced by the phosphoinositide 3-kinase complex II. Subsequently, vesicles pinch off from the tips of the tubular structures in a polarized manner, directly generating fragmentation products of the final size. This phase depends on the production of phosphatidylinositol-3,5-bisphosphate and the Fab1 complex. It is accelerated by the PI(3)P- and phosphatidylinositol 3,5-bisphosphate-binding protein Atg18p. Thus vacuoles fragment in two steps with distinct protein and lipid requirements.

Electrochemical synthesis of mesoporous CoPt nanowires for methanol oxidation

A new electrochemical method to synthesize mesoporous nanowires of alloys has been developed. Electrochemical deposition in ionic liquid-in-water (IL/W) microemulsion has been successful to grow mesoporous CoPt nanowires in the interior of polycarbonate membranes. The viscosity of the medium was high, but it did not avoid the entrance of the microemulsion in the interior of the membrane"s channels. The structure of the IL/W microemulsions, with droplets of ionic liquid (4 nm average diameter) dispersed in CoPt aqueous solution, defined the structure of the nanowires, with pores of a few nanometers, because CoPt alloy deposited only from the aqueous component of the microemulsion. The electrodeposition in IL/W microemulsion allows obtaining mesoporous structures in which the small pores must correspond to the size of the droplets of the electrolytic aqueous component of the microemulsion. The IL main phase is like a template for the confined electrodeposition. The comparison of the electrocatalytic behaviours towards methanol oxidation of mesoporous and compact CoPt nanowires of the same composition, demonstrated the porosity of the material. For the same material mass, the CoPt mesoporous nanowires present a surface area 16 times greater than compact ones, and comparable to that observed for commercial carbon-supported platinum nanoparticles.

Traitement percutané de la thrombose veineuse profonde aiguë [Catheter-based treatment for acute deep vein thrombosis].

Plus de la moitié des patients présentant une thrombose veineuse profonde des membres inférieurs développent un syndrome post-thrombotique. Le risque est particulièrement élevé en cas de thrombose de l'axe principal de drainage veineux comprenant la veine fémorale commune et les veines iliaques. Plusieurs études ont démontré que l'incidence du syndrome post-thrombotique peut être diminuée si une recanalisation des veines ilio-fémorales est obtenue dans la phase aiguë. A l'heure actuelle, des techniques de recanalisation percutanées sont proposées à des patients sélectionnés présentant une thrombose ilio-fémorale. Cet article a pour but de résumer les connaissances actuelles sur la recanalisation percutanée de la thrombose veineuse profonde aiguë. Nearly half of patients with acute lower limb deep vein thrombosis (DVT) develop a post-thrombotic syndrome (PTS). This risk is particularly high in case of proximal DVT of the common femoral and iliac vein, the major lower limbs venous outflow vessel. Several studies have demonstrated that PTS incidence can be reduced with early vein recanalisation. Currently, catheter-based recanalisation therapies can be offered to selected patients with acute ilio-femoral deep vein thrombosis. Aim of the present article is to summarize current knowledge on these catheter-based recanalisation therapies.

Active sulphide mine tailings impoundments as sources of contaminated drainage: controlling factors, methods of characterisation and geochemical constraints for mitigation

Low quality mine drainage from tailings facilities persists as one of the most significant global environmental concerns related to sulphide mining. Due to the large variation in geological and environmental conditions at mine sites, universal approaches to the management of mine drainage are not always applicable. Instead, site-specific knowledge of the geochemical behaviour of waste materials is required for the design and closure of the facilities. In this thesis, tailings-derived water contamination and factors causing the pollution were investigated in two coeval active sulphide mine sites in Finland: the Hitura Ni mine and the Luikonlahti Cu-Zn-Co-Ni mine and talc processing plant. A hydrogeochemical study was performed to characterise the tailingsderived water pollution at Hitura. Geochemical changes in the Hitura tailings were evaluated with a detailed mineralogical and geochemical investigation (solid-phase speciation, acid mine drainage potential, pore water chemistry) and using a spatial assessment to identify the mechanisms of water contamination. A similar spatial investigation, applying selective extractions, was carried out in the Luikonlahti tailings area for comparative purposes (Hitura low-sulphide tailings vs. Luikonlahti sulphide-rich tailings). At both sites, hydrogeochemistry of tailings seepage waters was further characterised to examine the net results of the processes observed within the impoundments and to identify constraints for water treatment. At Luikonlahti, annual and seasonal variation in effluent quality was evaluated based on a four-year monitoring period. Observations pertinent to future assessment and mine drainage prevention from existing and future tailings facilities were presented based on the results. A combination of hydrogeochemical approaches provided a means to delineate the tailings-derived neutral mine drainage at Hitura. Tailings effluents with elevated Ni, SO₄ ^2- and Fe content had dispersed to the surrounding aquifer through a levelled-out esker and underneath the seepage collection ditches. In future mines, this could be avoided with additional basal liners in tailings impoundments where the permeability of the underlying Quaternary deposits is inadequate, and with sufficiently deep ditches. Based on the studies, extensive sulphide oxidation with subsequent metal release may already initiate during active tailings disposal. The intensity and onset of oxidation depended on e.g. the Fe sulphide content of the tailings, water saturation level, and time of exposure of fresh sulphide grains. Continuous disposal decreased sulphide weathering in the surface of low-sulphide tailings, but oxidation initiated if they were left uncovered after disposal ceased. In the sulphide-rich tailings, delayed burial of the unsaturated tailings had resulted in thick oxidized layers, despite the continuous operation. Sulphide weathering and contaminant release occurred also in the border zones. Based on the results, the prevention of sulphide oxidation should already be considered in the planning of tailings disposal, taking into account the border zones. Moreover, even lowsulphide tailings should be covered without delay after active disposal ceases. The quality of tailings effluents showed wide variation within a single impoundment and between the two different types of tailings facilities assessed. The affecting factors included source materials, the intensity of weathering of tailings and embankment materials along the seepage flow path, inputs from the process waters, the water retention time in tailings, and climatic seasonality. In addition, modifications to the tailings impoundment may markedly change the effluent quality. The wide variation in the tailings effluent quality poses challenges for treatment design. The final decision on water management requires quantification of the spatial and seasonal fluctuation at the site, taking into account changes resulting from the eventual closure of the impoundment. Overall, comprehensive hydrogeochemical mapping was deemed essential in the identification of critical contaminants and their sources at mine sites. Mineralogical analysis, selective extractions, and pore water analysis were a good combination of methods for studying the weathering of tailings and in evaluating metal mobility from the facilities. Selective extractions with visual observations and pH measurements of tailings solids were, nevertheless, adequate in describing the spatial distribution of sulphide oxidation in tailings impoundments. Seepage water chemistry provided additional data on geochemical processes in tailings and was necessary for defining constraints for water treatment.

Cyclin-dependent kinases 4 and 6 control tumor progression and direct glucose oxidation in the pentose cycle

Cyclin-dependent kinases CDK4 and CDK6 are essential for the control of the cell cycle through the G1 phase. Aberrant expression of CDK4 and CDK6 is a hall- mark of cancer, which would suggest that CDK4 and CDK6 are attractive targets for cancer therapy. Herein, we report that calcein AM is a potent specific inhibitor of CDK4 and CDK6 in HCT116 human colon adenocarcinoma cells, inhibiting retinoblastoma protein (pRb) phosphorylation and inducing cell cycle arrest in the G1 phase. The metabolic effects of calcein AM (the calcein acetoxymethyl-ester) on HCT116 cells were also evaluated and the flux between the oxidative and non-oxidative branches of the pentose phos-phate pathway was significantly altered. To elucidate whe-ther these metabolic changes were due to the inhibition of CDK4 and CDK6, we also characterized the metabolic profile of a CDK4, CDK6 and CDK2 triple knockout of mouse embryonic fibroblasts. The results show that the metabolic profile associated with the depletion of CDK4, CDK6 and CDK2 coincides with the metabolic changes induced by calcein AM on HCT116 cells, thus confirming that the inhibition of CDK4 and CDK6 disrupts the balance between the oxidative and non-oxidative branches of the pentose phosphate pathway. Taken together, these results indicate that low doses of calcein can halt cell division and kill tumor cells. Thus, selective inhibition of CDK4 and CDK6 may be of greater pharmacological interest, since inhibitors of these kinases affect both cell cycle progression and the robust metabolic profile of tumors.

Wet oxidation of recalcitrant lignin waters: Experimental and kinetic studies

The dissertation is based on four articles dealing with recalcitrant lignin water purification. Lignin, a complicated substance and recalcitrant to most treatment technologies, inhibits seriously pulp and paper industry waste management. Therefore, lignin is studied, using WO as a process method for its degradation. A special attention is paid to the improvement in biodegradability and the reduction of lignin content, since they have special importance for any following biological treatment. In most cases wet oxidation is not used as a complete ' mineralization method but as a pre treatment in order to eliminate toxic components and to reduce the high level of organics produced. The combination of wet oxidation with a biological treatment can be a good option due to its effectiveness and its relatively low technology cost. The literature part gives an overview of Advanced Oxidation Processes (AOPs). A hot oxidation process, wet oxidation (WO), is investigated in detail and is the AOP process used in the research. The background and main principles of wet oxidation, its industrial applications, the combination of wet oxidation with other water treatment technologies, principal reactions in WO, and key aspects of modelling and reaction kinetics are presented. There is also given a wood composition and lignin characterization (chemical composition, structure and origin), lignin containing waters, lignin degradation and reuse possibilities, and purification practices for lignin containing waters. The aim of the research was to investigate the effect of the operating conditions of WO, such as temperature, partial pressure of oxygen, pH and initial concentration of wastewater, on the efficiency, and to enhance the process and estimate optimal conditions for WO of recalcitrant lignin waters. Two different waters are studied (a lignin water model solution and debarking water from paper industry) to give as appropriate conditions as possible. Due to the great importance of re using and minimizing the residues of industries, further research is carried out using residual ash of an Estonian power plant as a catalyst in wet oxidation of lignin-containing water. Developing a kinetic model that includes in the prediction such parameters as TOC gives the opportunity to estimate the amount of emerging inorganic substances (degradation rate of waste) and not only the decrease of COD and BOD. The degradation target compound, lignin is included into the model through its COD value (CODligning). Such a kinetic model can be valuable in developing WO treatment processes for lignin containing waters, or other wastewaters containing one or more target compounds. In the first article, wet oxidation of "pure" lignin water was investigated as a model case with the aim of degrading lignin and enhancing water biodegradability. The experiments were performed at various temperatures (110 -190°C), partial oxygen pressures (0.5 -1.5 MPa) and pH (5, 9 and 12). The experiments showed that increasing the temperature notably improved the processes efficiency. 75% lignin reduction was detected at the lowest temperature tested and lignin removal improved to 100% at 190°C. The effect of temperature on the COD removal rate was lower, but clearly detectable. 53% of organics were oxidized at 190°C. The effect of pH occurred mostly on lignin removal. Increasing the pH enhanced the lignin removal efficiency from 60% to nearly 100%. A good biodegradability ratio (over 0.5) was generally achieved. The aim of the second article was to develop a mathematical model for "pure" lignin wet oxidation using lumped characteristics of water (COD, BOD, TOC) and lignin concentration. The model agreed well with the experimental data (R2 = 0.93 at pH 5 and 12) and concentration changes during wet oxidation followed adequately the experimental results. The model also showed correctly the trend of biodegradability (BOD/COD) changes. In the third article, the purpose of the research was to estimate optimal conditions for wet oxidation (WO) of debarking water from the paper industry. The WO experiments were' performed at various temperatures, partial oxygen pressures and pH. The experiments showed that lignin degradation and organics removal are affected remarkably by temperature and pH. 78-97% lignin reduction was detected at different WO conditions. Initial pH 12 caused faster removal of tannins/lignin content; but initial pH 5 was more effective for removal of total organics, represented by COD and TOC. Most of the decrease in organic substances concentrations occurred in the first 60 minutes. The aim of the fourth article was to compare the behaviour of two reaction kinetic models, based on experiments of wet oxidation of industrial debarking water under different conditions. The simpler model took into account only the changes in COD, BOD and TOC; the advanced model was similar to the model used in the second article. Comparing the results of the models, the second model was found to be more suitable for describing the kinetics of wet oxidation of debarking water. The significance of the reactions involved was compared on the basis of the model: for instance, lignin degraded first to other chemically oxidizable compounds rather than directly to biodegradable products. Catalytic wet oxidation of lignin containing waters is briefly presented at the end of the dissertation. Two completely different catalysts were used: a commercial Pt catalyst and waste power plant ash. CWO showed good performance using 1 g/L of residual ash gave lignin removal of 86% and COD removal of 39% at 150°C (a lower temperature and pressure than with WO). It was noted that the ash catalyst caused a remarkable removal rate for lignin degradation already during the pre heating for `zero' time, 58% of lignin was degraded. In general, wet oxidation is not recommended for use as a complete mineralization method, but as a pre treatment phase to eliminate toxic or difficultly biodegradable components and to reduce the high level of organics. Biological treatment is an appropriate post treatment method since easily biodegradable organic matter remains after the WO process. The combination of wet oxidation with subsequent biological treatment can be an effective option for the treatment of lignin containing waters.

Behavior, performance and physiological parameters of pigs reared in deep bedding

An experiment was conducted to evaluate the behavior, performance and physiological parameters of pigs in different production systems. Twenty four animals in the growth phase were distributed in a randomized block design in three treatments: T1 - concrete floor, T2 - deep bedding with wood shaving, and T3 - deep bedding with coffee husks. The behavioral study was carried out by observing the animal behavior for an uninterrupted period of eight hours throughout seven weeks. The proportions of time spent in each behavior were characterized using the frequency histogram composition. Environmental (IBGTH), physiological (rectal and skin temperature and respiratory rate) and performance (weight gain, feed intake and feed conversion) parameters were measured in animals during the period. The production systems of deep bedding showed higher values of IBGTH. There was no effect of production systems evaluated on the performance parameters. Rectal temperature was higher in animals reared on deep bedding with coffee husks in relation to the concrete floor. The use of deep bedding benefited the behavior of piglets in the growth phase and it reduced the agonistic behavior among individuals.

Characterization and use of swine deep bedding ashes in cementitious composites

The deep bedding is a swine alternative production, especially in the finishing phase, whose byproduct can be recycled, reducing the environmental impact. The objectives of this study were to characterize the ash coming from the controlled burning of the swine deep bedding (SDBA) based on rice husk, and to evaluate their performance in composites as a partial substitute for Portland cement (PC). To measure the differences between SDBA and rice husk ash (RHA) as a reference, we have characterized: particle size distribution, real specific density, x-ray diffraction, electrical conductivity, scanning electron microscopy, chemical analysis and loss on ignition. Samples were prepared for two experimental series: control, and another one with the partial replacement of 30% of SDBA in relation to the mass of the Portland cement. According to the results obtained for physical and mechanical characterization, the composites with SDBA can be used as a constructive element in the rural construction.

Energetic efficiency of a deep bed swine production system

The aim of this work was to evaluate the energy flows of a commercial production system of swine deep bed in its finishing phase, located in Juiz de Fora, in the State of Minas Gerais, Brazil. Thus, an energy efficiency study was carried out by monitoring a lot of animals, during a 94-day period. The energy rate of each compound involved in the production process was quantified and the matrixes of energy consumption were determined in the form of animal feeding, electrical energy, piglets, material used as deep bed, human labor, equipment, swine buildings, production of alive swine for slaughter, organic fertilizer production (swine deep bed or swine deep litter). From the direct input energy, 80.57% correspond to animal feeding, 11.90% to pigs for slaughter and 6.76% to piglets, while from the energy output 53.45% correspond to the terminating swine and 46.55% to organic fertilizer (swine deep bed). By the results obtained, we can conclude that such production system has corresponded to an industrial and highly specialized agro ecosystem, importing a great part of the energy consumed in the production process, with 41% of energy efficiency.

Nitrogen fertilization by deep bedding swine production and its effects on dry matter production and accumulation of nutrients by maize

The goal of this study was to evaluate the nitrogen fertilization as deep litter for pigs in order to produce biomass and accumulate nutrients by the corn. A deep litter made of rice husk as organic compound, from a commercial pig farm during finishing phase, was used. After three consecutive batches of pigs, the deep litter was subjected to a maturation period of 50 days, and samples of this material were taken for analysis of agronomic value. The experimental design was completely randomized with five replicates. The treatments consisted of doses of 0, 75, 150 and 300mg dm-3 of N of deep litter, as well as an additional treatment with ammonium sulfate, with a dosage of 150mg dm-3 of N. After 45 days, corn plants were harvested in order to evaluate the total dry weight and nutrient concentrations of their aerial parts. Dry matter increases were found with more application of deep litter. Regarding control fertilization, the use of increasing dosages of deep litter allowed accumulation of K, reduced the availability of P, Ca, Mg, Zn and B and did not alter the concentrations of N, Cu, Fe and Mn.

Pulsed corona discharge as an advanced oxidation process for degradation of organic

Advanced oxidation processes (AOPs) have been studied and developed to suffice the effective removal of refractory and toxic compounds in polluted water. The quality and cost of wastewater treatment need improvements, and electric discharge technology has a potential to make a significant difference compared to other established AOPs based on energy efficiency. The generation of active oxidant species such as ozone and hydroxyl radicals by high voltage discharge is a relatively new technology for water treatment. Gas-phase pulsed corona discharge (PCD), where a treated aqueous solution is dispersed between corona-producing electrodes free of the dielectric barriers, was developed as an alternative approach to the problem. The short living radicals and ozone formed in the gas phase and at the gas-liquid interface react with dissolved impurities. PCD equipment has a relatively simple configuration, and with the reactor in an enclosed compartment, it is insensitive towards gas humidity and does not need the gas transport. In this thesis, PCD was used to study and evaluate the energy efficiency for degrading various organic compounds, as well as the chemistry of the oxidation products formed. The experiments investigate the aqueous oxidation of phenol, humic substances, pharmaceutical compounds (paracetamol, ibuprofen, indomethacin, salicylic acids, -estradiol), as well as lignin degradation and transformation to aldehydes. The study aims to establish the influence of initial concentration of the target pollutant, the pulsed discharge parameters, gas phase composition and the pH on the oxidation kinetics and the efficiency. Analytical methods to measure the concentrations of the target compounds and their by-products include HPLC, spectrophotometry, TOC and capillary electrophoresis. The results of the research included in this summary are presented in the attached publications and manuscripts accepted for publication. Pulsed corona discharge proved to be highly effective in oxidizing each of the target compounds, surpassing the closest competitor, conventional ozonation. The increase in oxidation efficiencies for some compounds in oxygen media and at lower pulse repetition frequencies shows a significant role of ozone. The role of the ·OH radicals was established in the surface reactions. The main oxidation products, formation of nitrates, and the lignin transformation were quantified. A compound specific approach is suggested for optimization of the PCD parameters that have the most significant impact on the oxidation energy efficiency because of the different characteristics and responses of the target compound to the oxidants, as well as different admixtures that are present in the wastewater. Further studies in the method’s safety (nitration and nitrosation of organic compounds, nitrite and nitrate formation enhancement) are needed for promoting the method.

Training deep convolutional architectures for vision

Les tâches de vision artificielle telles que la reconnaissance d’objets demeurent irrésolues à ce jour. Les algorithmes d’apprentissage tels que les Réseaux de Neurones Artificiels (RNA), représentent une approche prometteuse permettant d’apprendre des caractéristiques utiles pour ces tâches. Ce processus d’optimisation est néanmoins difficile. Les réseaux profonds à base de Machine de Boltzmann Restreintes (RBM) ont récemment été proposés afin de guider l’extraction de représentations intermédiaires, grâce à un algorithme d’apprentissage non-supervisé. Ce mémoire présente, par l’entremise de trois articles, des contributions à ce domaine de recherche. Le premier article traite de la RBM convolutionelle. L’usage de champs réceptifs locaux ainsi que le regroupement d’unités cachées en couches partageant les même paramètres, réduit considérablement le nombre de paramètres à apprendre et engendre des détecteurs de caractéristiques locaux et équivariant aux translations. Ceci mène à des modèles ayant une meilleure vraisemblance, comparativement aux RBMs entraînées sur des segments d’images. Le deuxième article est motivé par des découvertes récentes en neurosciences. Il analyse l’impact d’unités quadratiques sur des tâches de classification visuelles, ainsi que celui d’une nouvelle fonction d’activation. Nous observons que les RNAs à base d’unités quadratiques utilisant la fonction softsign, donnent de meilleures performances de généralisation. Le dernière article quand à lui, offre une vision critique des algorithmes populaires d’entraînement de RBMs. Nous montrons que l’algorithme de Divergence Contrastive (CD) et la CD Persistente ne sont pas robustes : tous deux nécessitent une surface d’énergie relativement plate afin que leur chaîne négative puisse mixer. La PCD à "poids rapides" contourne ce problème en perturbant légèrement le modèle, cependant, ceci génère des échantillons bruités. L’usage de chaînes tempérées dans la phase négative est une façon robuste d’adresser ces problèmes et mène à de meilleurs modèles génératifs.

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.

Singlet Oxygen in Microporous Silica Xerogel:Quantum Yield and Oxidation at the Gas–Solid Interface

The quantum yields of singlet oxygen production and lifetimes at the gas–solid interface in silica gel material are determined. Different photosensitizers (PS) are encapsulated in parallelepipedic xerogel monoliths (PS-SG). PS were chosen according to their known photooxidation properties: 9,10-dicyanoanthracene (DCA), 9,10-anthraquinone (ANT), and a benzophenone derivative, 4-benzoyl benzoic acid (4BB). These experiments are mainly based on time-resolved 1O2 phosphorescence detection, and the obtained FD and tD values are compared with those of a reference sensitizer for production, 1H-phenalen-1- one (PN), included in the same xerogel. The trend between their ability to oxidize organic pollutants in the gas phase and their efficiency for production is investigated through photooxidation experiments of a test pollutant dimethylsulfide (DMS). The FD value is high for DCA-SG relative to the PN reference, whereas it is slightly lower for 4BB-SG and for ANT-SG. FD is related to the production of sulfoxide and sulfone as the main oxidation products for DMS photosensitized oxidation. Additional mechanisms, leading to C!S bond cleaveage, appear to mainly occur for the less efficient singlet oxygen sensitizers 4BB-SG and ANTSG.

Oxidation kinetics of nickel nano crystallites obtained by controlled thermolysis of diaquabis(ethylenediamine) nickel(II) nitrate

The metal complex, [Ni(en)2(H2O)2](NO3)2 (en = ethylenediamine), was decomposed in a static furnace at 200 C by autogenous decomposition to obtain phase pure metallic nickel nanocrystallites. The nickel metal thus obtained was studied by XRD, IR spectra, SEM and CHN analysis. The nickel crystallites are in the nanometer range as indicated by XRD studies. The IR spectral studies and CHN analyses show that the surface is covered with a nitrogen containing species. Thermogravimetric mass gain shows that the product purity is high (93%). The formed nickel is stable and resistant to oxidation up to 350 C probably due to the coverage of nitrogen containing species. Activation energy for the oxidation of the prepared nickel nanocrystallites was determined by non-isothermal methods and was found to depend on the conversion ratio. The oxidation kinetics of the nickel crystallites obeyed a Johnson–Mehl–Avrami mechanism probably due to the special morphology and crystallite strain present on the metal.