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.

Influence of Angle of Air Injection and Particles in Bed Hydrodynamics of Swirling Fluidized Bed

The thesis presented here unveils an experimental study of the hydrodynamic characteristics of swirling fluidized bed viz. pressure drop across the distributor and the bed, minimum fluidizing velocity, bed behaviour and angle of air injection. In swirling fluidized bed the air is admitted to the bed at an angle 'Ѳ' to the horizontal. The vertical component of the velocity v sin Ѳ causes fluidization and the horizontal component v cos Ѳ contributes to swirl motion of the bed material.The study was conducted using spherical particles having sizes 3.2 mm, 5.5 mm & 7.4 mm as the bed materials. Each of these particles was made from high density polyethylene, nylon and acetal having relative densities of 0.93, 1.05 and 1.47 respectively.The experiments were conducted using conidour type distributors having four rows of slits. Altogether four distributors having angles of air injection (Φ)- 0°, 5°, 10° & 15° were designed and fabricated for the study. The total number of slits in each distributor was 144. The area of opening was 6220 mm2 making the percentage area of opening to 9.17. But the percentage useful area of opening of the distributor was 96.The experiments on the variation of distributor pressure drop with superficial velocity revealed that the distributor pressure drop decreases with angle of air injection. Investigations related to bed hydrodynamics were conducted using 2.5 kg of bed material. The bed pressure drop measurements were made along the radial direction of the distributor at distances of 60 mm, 90 mm, 120 mm & 150 mm from the centre of the distributor. It was noticed that after attaining minimum fluidizing velocity, the bed pressure drop increases along the radial direction of the distributor. But at a radial distance of 90 mm from the distributor centre, after attaining minimum fluidizing velocity the bed pressure drop remains almost constant. It was also observed that the bed pressure drop varies inversely with particle size as well as particle density.An attempt was made to determine the effect of various parameters on minimum fluidizing velocity. It was noticed that the minimum fluidizing velocity varies directly with angle of air injection (Φ), particle size and particle density.The study on the bed behaviour showed that the superficial velocity required for initiating various bed phenomena (such as swirl motion and separation of particles from the cone at the centre) increase with increase in particle size as well as particle density. It was also observed that the particle size and particle density directly influence the superficial velocity required for various regimes of bed behaviour such as linear variation of bed pressure drop, constant bed pressure drop and sudden increase or decrease in bed pressure drop.Experiments were also performed to study the effect of angle of air injection (Φ). It was noticed that the bed pressure drop decreases with angle of air injection. It was also noticed that the angle of air injection directly influence the superficial velocity required for initiating various bed phenomena as well as the various regimes of bed behaviour.

Biochemical & Pharmacological evaluation of liver oils of selected deep sea sharks and chimaeras of the Indian EEZ

With a seacoast of 8,1 18 km, an exclusive economic zone (EEZ) of 2 million square km, and with an area of about 30,000 square km under aquaculture, lndia produces close to six million tonnes of fish, over 4 per cent of the world fish production. While the marine waters upto 50m depth have been fully exploited, those beyond, remain unexplored. There is an ever increasing demand for fishery resources as food. The coastal fishery resources of the country are dwindling at a rapid pace and it becomes highly imperative that we search for alternate fishery resources for food. The option we have is to hunt for marine fishery resources. Studies pertaining to proximate composition, amino acid and fatty acid composition are essential to understand the nutraceutical values of these deep sea fishery resources. The present study was aimed to carry out proximate composition of deep sea fishery resources obtained during cruises onboard the FORV Sarise Sampada, to identify fishery resources which have appreciable lipid content and thereby analyse the bioactive potentials of marine lipids, to study the amino acid profile of these fishery resources, to understand the contents of SPA, MUFA and PUFA and to calculate the n3/n6 fatty acid contents. Though the presence of nutraceuticals was identified in the marine fishery resources their use as potential food resources deserve further investigation. So the study were carried out to calculate the hepatosomatic indices of sharks & chimaeras and conduct biochemical characterisation of liver oils of Apristurus indicus, Cenlrophorus scalprams, Centroselachus crepidater, Neoharriotta raleighana, and Harriotta pinnata obtained during cruises onboard the FORV Sugar Sampada.Therapeutic use of shark liver oil is evident from its use for centuries as a remedy to heal wounds and fight flu (Neil er al. 2006). Japanese seamen called it 'samedava' or "cure all". Shark liver oil is being promoted worldwide as a dietary supplement to boost the immune system, fight infections, to treat cancer and to lessen the side effects of conventional cancer treatment. These days more emphasis is laid on the nutritive benefits of shark liver oils especially on the omega 3 polyunsaturated fatty acids ( PUFAs) (Anandan er al. 2007) and alkylglycerols (AKGs) (Pugliese er al. I998) contained in them due to the high rise of inflammatory disorders such as arthritis, asthma and neurodegenerative diseases like Alzheimer’s, Parkinson’s and Schizophrenia. So the present study also evaluate the pharmacological properties with respect to analgesic, anti-inflammatory, anti pyretic and anti-ulcer effects of four different liver oils of sharks belonging to the Indian EEZ and to identify the components of oil responsible for these activities.The analgesic and anti-inflammatory activities of liver oils from Neoharriotra raleighana (NR), Centrosymnus crepidater (CC), Apristurus indicus (AI), and Centrophorus sculpratus (CS) sharks caught from the Arabian Sea and the Indian Ocean were compared. The main objectives also include determination of the cholesterol lowering effects of liver oils of Neoharriotra raleighana (NR) and Centrophorus sculpratus (CS) on the high fat diet induced dyslipidemia and to compare the impact of four isolipidemic diets, on levels of serum diagnostic marker enzymes, on lipid profile of blood and liver and antioxidant status of heart in male Albino rats. And also to study the efficacy of Centrophorus sculpratus (CS) liver oil against Complete Freund’s Adjuvant-induced arthritis and to compare the anti-inflammatory activity of this oil with a traditionally used anti-inflammatory substance gingerol (oleoresin extracted from ginger.). The results of the present study indicated that both (Centrophorus sculpratus liver oils as well as gingerol extracts proved to be effective natural remedies against CFA-induced arthritis in Albino rats.

Dynamic Injection of Scribble Features into Graphical Diagram Editors

Almost everyone sketches. People use sketches day in and day out in many different and heterogeneous fields, to share their thoughts and clarify ambiguous interpretations, for example. The media used to sketch varies from analog tools like flipcharts to digital tools like smartboards. Whereas analog tools are usually affected by insufficient editing capabilities like cut/copy/paste, digital tools greatly support these scenarios. Digital tools can be grouped into informal and formal tools. Informal tools can be understood as simple drawing environments, whereas formal tools offer sophisticated support to create, optimize and validate diagrams of a certain application domain. Most digital formal tools force users to stick to a concrete syntax and editing workflow, limiting the user’s creativity. For that reason, a lot of people first sketch their ideas using the flexibility of analog or digital informal tools. Subsequently, the sketch is "portrayed" in an appropriate digital formal tool. This work presents Scribble, a highly configurable and extensible sketching framework which allows to dynamically inject sketching features into existing graphical diagram editors, based on Eclipse GEF. This allows to combine the flexibility of informal tools with the power of formal tools without any effort. No additional code is required to augment a GEF editor with sophisticated sketching features. Scribble recognizes drawn elements as well as handwritten text and automatically generates the corresponding domain elements. A local training data library is created dynamically by incrementally learning shapes, drawn by the user. Training data can be shared with others using the WebScribble web application which has been created as part of this work.

Consistent approximate models of the global atmosphere: shallow, deep, hydrostatic, quasi-hydrostatic and non-hydrostatic

We study global atmosphere models that are at least as accurate as the hydrostatic primitive equations (HPEs), reviewing known results and reporting some new ones. The HPEs make spherical geopotential and shallow atmosphere approximations in addition to the hydrostatic approximation. As is well known, a consistent application of the shallow atmosphere approximation requires omission of those Coriolis terms that vary as the cosine of latitude and of certain other terms in the components of the momentum equation. An approximate model is here regarded as consistent if it formally preserves conservation principles for axial angular momentum, energy and potential vorticity, and (following R. Müller) if its momentum component equations have Lagrange's form. Within these criteria, four consistent approximate global models, including the HPEs themselves, are identified in a height-coordinate framework. The four models, each of which includes the spherical geopotential approximation, correspond to whether the shallow atmosphere and hydrostatic (or quasi-hydrostatic) approximations are individually made or not made. Restrictions on representing the spatial variation of apparent gravity occur. Solution methods and the situation in a pressure-coordinate framework are discussed. © Crown copyright 2005.

Initiation of deep convection at marginal instability in an ensemble of mesoscale models: a case-study from COPS

The present study investigates the initiation of precipitating deep convection in an ensemble of convection-resolving mesoscale models. Results of eight different model runs from five non-hydrostatic models are compared for a case of the Convective and Orographically-induced Precipitation Study (COPS). An isolated convective cell initiated east of the Black Forest crest in southwest Germany, although convective available potential energy was only moderate and convective inhibition was high. Measurements revealed that, due to the absence of synoptic forcing, convection was initiated by local processes related to the orography. In particular, the lifting by low-level convergence in the planetary boundary layer is assumed to be the dominant process on that day. The models used different configurations as well as different initial and boundary conditions. By comparing the different model performance with each other and with measurements, the processes which need to be well represented to initiate convection at the right place and time are discussed. Besides an accurate specification of the thermodynamic and kinematic fields, the results highlight the role of boundary-layer convergence features for quantitative precipitation forecasts in mountainous terrain.

Cloud-resolving simulations of deep convection over a heated mountain

Cloud-resolving numerical simulations of airflow over a diurnally heated mountain ridge are conducted to explore the mechanisms and sensitivities of convective initiation under high pressure conditions. The simulations are based on a well-observed convection event from the Convective and Orographically Induced Precipitation Study (COPS) during summer 2007, where an isolated afternoon thunderstorm developed over the Black Forest mountains of central Europe, but they are idealized to facilitate understanding and reduce computational expense. In the conditionally unstable but strongly inhibited flow under consideration, sharp horizontal convergence over the mountain acts to locally weaken the inhibition and moisten the dry midtroposphere through shallow cumulus detrainment. The onset of deep convection occurs not through the deep ascent of a single updraft but rather through a rapid succession of thermals that are vented through the mountain convergence zone into the deepening cloud mass. Emerging thermals rise through the saturated wakes of their predecessors, which diminishes the suppressive effects of entrainment and allows for rapid glaciation above the freezing level as supercooled cloud drops rime onto preexisting ice particles. These effects strongly enhance the midlevel cloud buoyancy and enable rapid ascent to the tropopause. The existence and vigor of the convection is highly sensitive to small changes in background wind speed U0, which controls the strength of the mountain convergence and the ability of midlevel moisture to accumulate above the mountain. Whereas vigorous deep convection develops for U0 = 0 m s−1, deep convection is completely eliminated for U0 = 3 m s−1. Although deep convection is able to develop under intermediate winds (U0 = 1.5 m s−1), its formation is highly sensitive to small-amplitude perturbations in the initial flow.

Importance of density-compensated temperature change for deep North Atlantic Ocean heat uptake

The efficiency with which the oceans take up heat has a significant influence on the rate of global warming. Warming of the ocean above 700 m over the past few decades has been well documented. However, most of the ocean lies below 700 m. Here we analyse observations of heat uptake into the deep North Atlantic. We find that the extratropical North Atlantic as a whole warmed by 1.45±0.5×1022 J between 1955 and 2005, but Lower North Atlantic Deep Water cooled, most likely as an adjustment from an early twentieth-century warm period. In contrast, the heat content of Upper North Atlantic Deep Water exhibited strong decadal variability. We demonstrate and quantify the importance of density-compensated temperature anomalies for long-term heat uptake into the deep North Atlantic. These anomalies form in the subpolar gyre and propagate equatorwards. High salinity in the subpolar gyre is a key requirement for this mechanism. In the past 50 years, suitable conditions have occurred only twice: first during the 1960s and again during the past decade. We conclude that heat uptake through density-compensated temperature anomalies will contribute to deep ocean heat uptake in the near term. In the longer term, the importance of this mechanism will be determined by competition between the multiple processes that influence subpolar gyre salinity in a changing climate.

The energy-cycle analysis of the interactions between shallow and deep atmospheric convection

Interactions between different convection modes can be investigated using an energy–cycle description under a framework of mass–flux parameterization. The present paper systematically investigates this system by taking a limit of two modes: shallow and deep convection. Shallow convection destabilizes itself as well as the other convective modes by moistening and cooling the environment, whereas deep convection stabilizes itself as well as the other modes by drying and warming the environment. As a result, shallow convection leads to a runaway growth process in its stand–alone mode, whereas deep convection simply damps out. Interaction between these two convective modes becomes a rich problem, even when it is limited to the case with no large–scale forcing, because of these opposing tendencies. Only if the two modes are coupled at a proper level can a self–sustaining system arise, exhibiting a periodic cycle. The present study establishes the conditions for self–sustaining periodic solutions. It carefully documents the behaviour of the two mode system in order to facilitate the interpretation of global model behaviours when this energy–cycle is implemented as a closure into a convection parameterization in future.

In at the Deep End. Impressions of farming in a year gaining experience: 1952-3; 1978-85

The book describes a wide variety of students’ experiences in their practical year prior to entering University to study BSc Agriculture. Until comparatively recently it was the normal requirement for all such students, whether or not they already had home farming experience, to gain a full year’s experience of practical agriculture – and to write a report thereon. This record of 41 students’ reports of the pre-entry year begins with Paul’s own experience in the early 1950s before 41 reports from 30 or more years ago. The essays provide compelling and fascinating stories, well-articulated with clear acknowledgement for most part of the humanity and the warmth with which each student was treated by farmers and farm workers alike, despite the difference in both age and experience (considerable!). [This summary is an extract from the full overview which is archived here together with the book.]

Simulated last glacial maximum D14Catm and the deep glacial ocean carbon reservoir

∆14Catm has been estimated as 420 ± 80‰ (IntCal09) during the Last Glacial Maximum (LGM) compared to preindustrial times (0‰), but mechanisms explaining this difference are not yet resolved. ∆14Catm is a function of both cosmogenic production in the high atmosphere and of carbon cycling and partitioning in the Earth system. 10Be-based reconstructions show a contribution of the cosmogenic production term of only 200 ± 200‰ in the LGM. The remaining 220‰ have thus to be explained by changes in the carbon cycle. Recently, Bouttes et al. (2010, 2011) proposed to explain most of the difference in pCO2atm and δ13C between glacial and interglacial times as a result of brine-induced ocean stratification in the Southern Ocean. This mechanism involves the formation of very saline water masses that contribute to high carbon storage in the deep ocean. During glacial times, the sinking of brines is enhanced and more carbon is stored in the deep ocean, lowering pCO2atm. Moreover, the sinking of brines induces increased stratification in the Southern Ocean, which keeps the deep ocean well isolated from the surface. Such an isolated ocean reservoir would be characterized by a low ∆14C signature. Evidence of such 14C-depleted deep waters during the LGM has recently been found in the Southern Ocean (Skinner et al. 2010). The degassing of this carbon with low ∆14C would then reduce ∆14Catm throughout the deglaciation. We have further developed the CLIMBER-2 model to include a cosmogenic production of 14C as well as an interactive atmospheric 14C reservoir. We investigate the role of both the sinking of brine and cosmogenic production, alongside iron fertilization mechanisms, to explain changes in ∆14Catm during the last deglaciation. In our simulations, not only is the sinking of brine mechanism consistent with past ∆14C data, but it also explains most of the differences in pCO2atm and ∆14Catm between the LGM and preindustrial times. Finally, this study represents the first time to our knowledge that a model experiment explains glacial-interglacial differences in pCO2atm, δ13C, and ∆14C together with a coherent LGM climate.

Seismic activity triggered by water wells in the Parana Basin, Brazil

Triggered seismicity is commonly associated with deep water reservoirs or injection wells where water is injected at high pressure into the reservoir rock. However, earth tremors related solely to the opening of groundwater wells are extremely rare. Here we present a clear case of seismicity induced by pore-pressure changes following the drilling of water wells that exploit a confined aquifer in the intracratonic Parana Basin of southeastern Brazil. Since 2004, shallow seismic activity, with magnitudes up to 2.9 and intensities V MM, has been observed near deep wells (120-200 m) that were drilled in early 2003 near the town of Bebedouro. The wells were drilled for irrigation purposes, cross a sandstone layer about 60-80 m thick and extract water from a confined aquifer in fractured zones between basalt flow layers. Seismic activity, mainly event swarms, has occurred yearly since 2004, mostly during the rainy season when the wells are not pumped. During the dry season when the wells are pumped almost continuously, the activity is very low. A seismographic network, installed in March 2005, has located more than 2000 microearthquakes. The events are less than 1 km deep (mostly within the 0.5 km thick basalt layer) and cover an area roughly 1.5 km x 5 km across. The seismicity generally starts in a small area and expands to larger distances with an equivalent hydraulic diffusivity ranging from 0.06 to 0.6 m(2)/s. Geophysical and geothermal logging of several wells in the area showed that water from the shallow sandstone aquifer enters the well at the top and usually forms waterfalls. The waterfalls flow down the sides of the wells and feed the confined, fractured aquifer in the basalt layer at the bottom. Two seismic areas are observed: the main area surrounds several wells that are pumped continuously during the dry season, and a second area near another well (about 10 km from the first area) that is not used for irrigation and not pumped regularly. The main area displays cyclic annual activity, but the second area does not. We explain the earthquake swarms as being triggered by pore pressure diffusion in the fractured basalt layer due to additional pressure from the newly connected surface aquifer. This reaches critically prestressed areas up to a few kilometers away from the wells. During periods of continuous pumping, the reduction of pore pressure in the confined aquifer stops the seismic activity. Our study suggests that this kind of activity may be more common than previously thought and implies that many other cases of small tremors associated with the drilling of water wells may have gone unnoticed.

Short- and long-term anxiogenic effects induced by a single injection of subconvulsant doses of pilocarpine in rats: investigation of the putative role of hippocampal pathways

Behavioral consequences of convulsive episodes are well documented, but less attention was paid to changes that occur in response to subconvulsant doses of drugs. We investigated short- and long-term effects of a single systemic injection of a subconvulsant dose of pilocarpine on the behavior of rats as evaluated in the elevated plus maze. Pilocarpine induced an anxiogenic-like profile 24 h later, and this effect persisted for up to 3 months (% of time spent on open arms at 24 h, control = 35.47 +/- 3.23; pilocarpine 150 = 8.2 +/- 2.6; 3 months, control = 31.9 +/- 5.5; pilocarpine 150 = 9.3 +/- 4.9). Temporary inactivation of fimbria-fornix with lidocaine 4% promoted an anxiolytic-like effect per se, suggesting a tonic control of this pathway on the modulation of anxiety-related behaviors. Lidocaine also reduced the anxiogenic-like profile of animals tested 1 month after pilocarpine treatment (% of time spent on open arms, saline + phosphate-buffered saline (PBS) = 31.7 + 3.7; saline + lidocaine = 54.4 + 4.7; pilocarpine + PBS = 10.3 + 4.1; pilocarpine + lidocaine = 40.1 + 9.1). To determine whether the anxiogenic-like effect was mediated by septal region or by direct hippocampal projections to the diencephalon, the neural transmission of post-commissural fornix was blocked, and a similar reduction in the anxiogenic-like effect of pilocarpine was observed. Our findings suggest that a single systemic injection of pilocarpine may induce long-lasting anxiogenic-like behavior in rats, an effect that appears to be mediated, in part, through a direct path from hippocampus to medial hypothalamic sites involved in fear responses.

Development of a sequential injection chromatography (SIC) method for determination of simazine, atrazine, and propazine

This paper describes the development of a sequential injection chromatography (SIC) procedure for separation and quantification of the herbicides simazine, atrazine, and propazine exploring the low backpressure of a 2.5 cm long monolithic C(18) column. The separation of the three compounds was achieved in less than 90 s with resolution > 1.5 using a mobile phase composed by ACN/1.25 mmol/L acetate buffer (pH 4.5) at the volumetric ratio of 35:65 and flow rate of 40 mu L/s. Detection was made at 223 nm using a flow cell with 40 mm of optical path length. The LOD was 10 mu g/L for the three triazines and the quantification limits were of 30 mu g/L for simazine and propazine and 40 mu g/L for atrazine. The sampling frequency is 27 samples per hour, consuming 1.1 mL of ACN per analysis. The proposed methodology was applied to spiked water samples and no statistically significant differences were observed in comparison to a conventional HPLC-UV method. The major metabolites of atrazine and other herbicides did not interfere in the analysis, being eluted from the column either together with the unretained peak, or at retention times well-resolved from the studied compounds.

Fast batch injection analysis of H(2)O(2) using an array of Pt-modified gold microelectrodes obtained from split electronic chips

A fast and robust analytical method for amperometric determination of hydrogen peroxide (H(2)O(2)) based on batch injection analysis (BIA) on an array of gold microelectrodes modified with platinum is proposed. The gold microelectrode array (n = 14) was obtained from electronic chips developed for surface mounted device technology (SMD), whose size offers advantages to adapt them in batch cells. The effect of the dispensing rate, volume injected, distance between the platinum microelectrodes and the pipette tip, as well as the volume of solution in the cell on the analytical response were evaluated. The method allows the H(2)O(2) amperometric determination in the concentration range from 0.8 mu mol L(-1) to 100 mu mol L(-1). The analytical frequency can attain 300 determinations per hour and the detection limit was estimated in 0.34 mu mol L(-1) (3 sigma). The anodic current peaks obtained after a series of 23 successive injections of 50 mu L of 25 mu mol L(-1) H(2)O(2) showed an RSD < 0.9%. To ensure the good selectivity to detect H(2)O(2), its determination was performed in a differential mode, with selective destruction of the H(2)O(2) with catalase in 10 mmol L(-1) phosphate buffer solution. Practical application of the analytical procedure involved H(2)O(2) determination in rainwater of Sao Paulo City. A comparison of the results obtained by the proposed ampermetric method with another one which combines flow injection analysis (FIA) with spectrophotometric detection showed good agreement. (C) 2011 Elsevier B.V. All rights reserved.