Performance of the Bayesian online algorithm for the perceptron

In this letter, we derive continuum equations for the generalization error of the Bayesian online algorithm (BOnA) for the one-layer perceptron with a spherical covariance matrix using the Rosenblatt potential and show, by numerical calculations, that the asymptotic performance of the algorithm is the same as the one for the optimal algorithm found by means of variational methods with the added advantage that the BOnA does not use any inaccessible information during learning. © 2007 IEEE.

Évaluation de la modélisation de la taille de grain de neige du modèle multi-couches thermodynamique SNOWPACK: implication dans l'évaluation des risques d'avalanches

Résumé: L’Institut pour l'étude de la neige et des avalanches en Suisse (SLF) a développé SNOWPACK, un modèle thermodynamique multi-couches de neige permettant de simuler les propriétés géophysiques du manteau neigeux (densité, température, taille de grain, teneur en eau, etc.) à partir desquelles un indice de stabilité est calculé. Il a été démontré qu’un ajustement de la microstructure serait nécessaire pour une implantation au Canada. L'objectif principal de la présente étude est de permettre au modèle SNOWPACK de modéliser de manière plus réaliste la taille de grain de neige et ainsi obtenir une prédiction plus précise de la stabilité du manteau neigeux à l’aide de l’indice basé sur la taille de grain, le Structural Stability Index (SSI). Pour ce faire, l’erreur modélisée (biais) par le modèle a été analysée à l’aide de données précises sur le terrain de la taille de grain à l’aide de l’instrument IRIS (InfraRed Integrated Sphere). Les données ont été recueillies durant l’hiver 2014 à deux sites différents au Canada : parc National des Glaciers, en Colombie-Britannique ainsi qu’au parc National de Jasper. Le site de Fidelity était généralement soumis à un métamorphisme à l'équilibre tandis que celui de Jasper à un métamorphisme cinétique plus prononcé. Sur chacun des sites, la stratigraphie des profils de densités ainsi des profils de taille de grain (IRIS) ont été complétés. Les profils de Fidelity ont été complétés avec des mesures de micropénétromètre (SMP). L’analyse des profils de densité a démontré une bonne concordance avec les densités modélisées (R[indice supérieur 2]=0.76) et donc la résistance simulée pour le SSI a été jugée adéquate. Les couches d’instabilités prédites par SNOWPACK ont été identifiées à l’aide de la variation de la résistance dans les mesures de SMP. L’analyse de la taille de grain optique a révélé une surestimation systématique du modèle ce qui est en accord avec la littérature. L’erreur de taille de grain optique dans un environnement à l’équilibre était assez constante tandis que l’erreur en milieux cinétique était plus variable. Finalement, une approche orientée sur le type de climat représenterait le meilleur moyen pour effectuer une correction de la taille de grain pour une évaluation de la stabilité au Canada.

Development and evaluation of neural network freeway incident detection models using field data

This paper discusses a multi-layer feedforward (MLF) neural network incident detection model that was developed and evaluated using field data. In contrast to published neural network incident detection models which relied on simulated or limited field data for model development and testing, the model described in this paper was trained and tested on a real-world data set of 100 incidents. The model uses speed, flow and occupancy data measured at dual stations, averaged across all lanes and only from time interval t. The off-line performance of the model is reported under both incident and non-incident conditions. The incident detection performance of the model is reported based on a validation-test data set of 40 incidents that were independent of the 60 incidents used for training. The false alarm rates of the model are evaluated based on non-incident data that were collected from a freeway section which was video-taped for a period of 33 days. A comparative evaluation between the neural network model and the incident detection model in operation on Melbourne's freeways is also presented. The results of the comparative performance evaluation clearly demonstrate the substantial improvement in incident detection performance obtained by the neural network model. The paper also presents additional results that demonstrate how improvements in model performance can be achieved using variable decision thresholds. Finally, the model's fault-tolerance under conditions of corrupt or missing data is investigated and the impact of loop detector failure/malfunction on the performance of the trained model is evaluated and discussed. The results presented in this paper provide a comprehensive evaluation of the developed model and confirm that neural network models can provide fast and reliable incident detection on freeways. (C) 1997 Elsevier Science Ltd. All rights reserved.

Australia and New Zealand satellite coverage using a microstrip patch reflectarray

This paper presents the design of Ku-band (12.25 12.75 GHz) dual-polarized reflectarrays for Optus BI satellites to obtain a contoured beam for Australia and New Zealand. The specified radiation pattern is synthesized using a phase-only synthesis method based on the concept of intersection approach. Having determined the phasing data, single- and double-layer reflectarrays ore designed using variable-size rectangular patches. The performances of the two reflectarrays are assessed by comparing their radiation patterns with the assumed pattern. (C) 2003 Wiley Periodicals, Inc. Microwave Opt Technol Lett 37: 321-325, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10. 1002/mop. 10907.

A unit cell waveguide model of a reflectarray formed by microstrip patches and slots

An equivalent unit cell waveguide approach (WGA) is described to obtain reflection coefficient phase curves for designing a microstrip patch reflectarray supported by a ground plane with periodic apertures or slots. Based on the presented theory, a computer algorithm for determining the reflection coefficient of a plane wave normally incident on a multi-layer structure of patches and apertures is developed. The validity of the developed algorithm is verified by comparing the obtained results with those published in the literature and the ones generated by Agilent High Frequency Structure Simulator (HFSS). A good agreement in all the presented examples is obtained, proving that the developed theory and computer algorithm can be an effective tool for designing multi-layer microstrip reflectarrays with a periodically perforated ground plane. (C) 2003 Wiley Periodicals, Inc.

A brief survey on online analysis of movement data

The increasing availability of mobility data and the awareness of its importance and value have been motivating many researchers to the development of models and tools for analyzing movement data. This paper presents a brief survey of significant research works about modeling, processing and visualization of data about moving objects. We identified some key research fields that will provide better features for online analysis of movement data. As result of the literature review, we suggest a generic multi-layer architecture for the development of an online analysis processing software tool, which will be used for the definition of the future work of our team.

Colour filtering in a-SiC : H based p-i-n-p-i-n cells: A trade-off between bias polarity and absorption regions

A large area colour imager optically addressed is presented. The colour imager consists of a thin wide band gap p-i-n a-SiC:H filtering element deposited on the top of a thick large area a-SiC:H(-p)/a-Si:H(-i)/a-SiC:H(-n) image sensor, which reveals itself an intrinsic colour filter. In order to tune the external applied voltage for full colour discrimination the photocurrent generated by a modulated red light is measured under different optical and electrical bias. Results reveal that the integrated device behaves itself as an imager and a filter giving information not only on the position where the optical image is absorbed but also on it wavelength and intensity. The amplitude and sign of the image signals are electrically tuneable. In a wide range of incident fluxes and under reverse bias, the red and blue image signals are opposite in sign and the green signal is suppressed allowing blue and red colour recognition. The green information is obtained under forward bias, where the blue signal goes down to zero and the red and green remain constant. Combining the information obtained at this two applied voltages a RGB colour image picture can be acquired without the need of the usual colour filters or pixel architecture. A numerical simulation supports the colour filter analysis.

Dimensionless analysis of constrained damping treatments

One of the most effective ways of controlling vibrations in plate or beam structures is by means of constrained viscoelastic damping treatments. Contrary to the unconstrained configuration, the design of constrained and integrated layer damping treatments is multifaceted because the thickness of the viscoelastic layer acts distinctly on the two main counterparts of the strain energy the volume of viscoelastic material and the shear strain field. In this work, a parametric study is performed exploring the effect that the design parameters, namely the thickness/length ratio, constraining layer thickness, material modulus, natural mode and boundary conditions have on these two counterparts and subsequently, on the treatment efficiency. This paper presents five parametric studies, namely, the thickness/length ratio, the constraining layer thickness, material properties, natural mode and boundary conditions. The results obtained evidence an interesting effect when dealing with very thin viscoelastic layers that contradicts the standard treatment efficiency vs. layer thickness relation; hence, the potential optimisation of constrained and integrated viscoelastic treatments through the use of properly designed thin multilayer configurations is justified. This work presents a dimensionless analysis and provides useful general guidelines for the efficient design of constrained and integrated damping treatments based on single or multi-layer configurations. (C) 2012 Elsevier Ltd. All rights reserved.

Analysis and simulation of a-Si Ha-SiC HPINIP structures for color image detection

It is presented in this paper a study on the photo-electronic properties of multi layer a-Si: H/a-SiC: H p-i-n-i-p structures. This study is aimed to give an insight into the internal electrical characteristics of such a structure in thermal equilibrium, under applied Was and under different illumination condition. Taking advantage of this insight it is possible to establish a relation among-the electrical behavior of the structure the structure geometry (i.e. thickness of the light absorbing intrinsic layers and of the internal n-layer) and the composition of the layers (i.e. optical bandgap controlled through percentage of carbon dilution in the a-Si1-xCx: H layers). Showing an optical gain for low incident light power controllable by means of externally applied bias or structure composition, these structures are quite attractive for photo-sensing device applications, like color sensors and large area color image detector. An analysis based on numerical ASCA simulations is presented for describing the behavior of different configurations of the device and compared with experimental measurements (spectral response and current-voltage characteristic). (c) 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Controlo de águas do processo e optimização da produção de papel

Na Fábrica de Papel da Ponte Redonda fabricam-se sacos de papel multi-folhas e papel reciclado do tipo Kraft. Tendo em consideração a primeira actividade, é de grande interesse optimizar o processo de fabrico de papel com vista a incorporara a máxima taxa de papel produzido internamente nas diferentes camadas dos sacos de papel. Os papéis de maior interesse são os do tipo Fluting e Liners, tendo sido produzidos em 2010 um total de 4,9 mil toneladas, ou seja 90% de todo o papel fabricado em 2010, correspondendo a a 4 mil toneladas de papéis do tipo Liners e 0,9 mil toneladas para os papéis do tipo Fluting. Nos papéis do tipo Liners incluem-se os papéis do tipo Test-Liner e Kraft-Liner, representando em termos produtivos valores idênticos. No âmbito deste trabalho, em que se pretendeu controlar as águas do processo e optimizar a produção de papel, foram introduzidos uma unidade de flutuação e um sistema que permitisse regular a consistência da suspensão fibrosa à entrada da máquina do papel, e foram ainda estudadas as possibilidades de adição de produtos químicos para melhorar as características da pasta assim como um tratamento microbiológico mais eficaz para todo o processo. Para se avaliar se as medidas implementadas teriam um impacto positivo na qualidade desses dois tipos de papéis, desenvolveu-se o trabalho em duas fases: a primeira envolve a introdução de um sistema de flutuação e de um sistema de controlo de consistência da pasta, assim como a selecção de produtos químicos a adicionar ao processo. A segunda fase consistiu na avaliação do efeito destas medidas nas características do papel fabricado. Para o efeito foram escolhidos dois tipos de papel e de diferentes gramagens, nomeadamente Test-Liner de 80 g/m2 e Fluting de 110 g/m2. Introduziu-se um flutuador com o objectivo de tratar parte das águas do processo de fabrico com vista a serem reutilizadas em determinadas aplicações possíveis para a qualidade da água obtida (lavagens e água do processo), de modo a conseguir-se uma poupança de água, assim como aproveitar-se as lamas resultantes, ricas em fibra de celulose, para utilizá-las como matéria-prima. Foi introduzido um regulador de consistência no processo da Ponte Redonda com o objectivo de alimentar de uma forma constante a consistência da pasta à entrada da máquina do papel proporcionando uma melhor formação da folha, devido à ligação entre fibras, na direcção máquina e direcção transversal. Esse sistema inovador é um Regulador de Consistência que vem proporcionar à máquina do papel uma alimentação em fibra mais constante. O fabrico de papel apenas a partir de fibras de celulose não permitirá obter um papel com as características desejadas para a sua utilização. Para corrigir estas deficiências, são adicionados produtos químicos para atribuir ou melhorar as propriedades dos papéis. Desta forma considerou-se interessante introduzir no processo um agente de retenção numa fase posterior à preparação da pasta e antes da chegada à máquina de papel, de forma a melhorar as características da suspensão fibrosa. Assim foi implementado um sistema cuja eficácia foi avaliada. Concluiu-se que com a sua implementação a máquina de papel apresentou melhores resultados na drenagem e na turbidez da água removida, significando uma água com menor teor de matéria suspensa e dissolvida, devido à melhor agregação das fibras dispersas na suspensão fibrosa, levando a um aumento da drenagem e consequentemente melhor eficiência das prensas e secaria. Foi também elaborado um estudo para introdução de um sistema de tratamento microbiológico em todo o processo de fabrico de papel, devido à existência de microorganismos prejudiciais ao seu fabrico. Concluiu-se que a água clarificada proveniente do flutuador apresentou qualidade aceitável para os objectivos pretendidos. No entanto, considerando a eficiência de 26,5% na remoção de sólidos suspensos será necessário mais algum tempo de utilização da água clarificada, cerca de um ano, para avaliar se esta terá algum efeito prejudicial nos equipamentos. Verificou-se que devido à existência de microrganismos em todo o processo de fabrico de papel será necessário efectuar lavagens aos tinões, tanques e circuitos com alguma regularidade, aproveitando-se as paragens do processo assim como implementar um sistema de tratamento microbiológico mais eficaz. Em resultado das medidas implementadas concluiu-se que os papéis produzidos apresentaram melhorias, tendo-se obtido melhores resultados em todos os ensaios de resistência. No papel do tipo Test-Liner destacam-se os bons resultados nos ensaios de superfície, Cobb60 e rebentamento. No caso do parâmetro do Cobb60, foi um resultado surpreendente visto que por norma este tipo de papéis reciclados não suportam este ensaio. Concluiu-se também que as medidas implementadas proporcionaram uma melhor agregação e ligação entre fibras, e melhor formação da folha na máquina do papel proporcionando aos papéis propriedades físico-mecânicas mais interessantes.

Target-specific multiphysics modeling for thermal medicine applications

Dissertation to obtain the degree of Doctor of Philosophy in Biomedical Engineering

J2EE. Diseño e implementación de un Framework de Persistencia

Anàlisi i desenvolupament d'un conjunt de components que simplifiquin i agilitin el desenvolupament de la capa de persistència en aplicacions multicapa, especialment per a les aplicacions desenvolupades amb J2EE.

Web-CRM en J2EE

Crear una aplicació i-CRM, el Customer Relationship Management és un sistema que permet a les empreses manejar tota la informació relativa als seus clients des d'una única plataforma de programari. L'esmentada aplicació tindrà una interfície Web, i una estructura multicapa, que pugui ser emprada des de qualsevol localització i el nombre d'usuaris no sigui una limitació. Bàsicament, consistirà en una aplicació on es puguin crear entitats comercials (clients i potencials), i mantenir la relació bàsica amb els mateixos, aquesta es restringirà als contactes comercials, incidències i campanyes de màrqueting relacional.

Artificial neural network study of whole-cell bacterial bioreporter response determined using fluorescence flow cytometry.

Genetically engineered bioreporters are an excellent complement to traditional methods of chemical analysis. The application of fluorescence flow cytometry to detection of bioreporter response enables rapid and efficient characterization of bacterial bioreporter population response on a single-cell basis. In the present study, intrapopulation response variability was used to obtain higher analytical sensitivity and precision. We have analyzed flow cytometric data for an arsenic-sensitive bacterial bioreporter using an artificial neural network-based adaptive clustering approach (a single-layer perceptron model). Results for this approach are far superior to other methods that we have applied to this fluorescent bioreporter (e.g., the arsenic detection limit is 0.01 microM, substantially lower than for other detection methods/algorithms). The approach is highly efficient computationally and can be implemented on a real-time basis, thus having potential for future development of high-throughput screening applications.

Efficient learning of reactive robot behaviors with a Neural-Q_learning approach

The purpose of this paper is to propose a Neural-Q_learning approach designed for online learning of simple and reactive robot behaviors. In this approach, the Q_function is generalized by a multi-layer neural network allowing the use of continuous states and actions. The algorithm uses a database of the most recent learning samples to accelerate and guarantee the convergence. Each Neural-Q_learning function represents an independent, reactive and adaptive behavior which maps sensorial states to robot control actions. A group of these behaviors constitutes a reactive control scheme designed to fulfill simple missions. The paper centers on the description of the Neural-Q_learning based behaviors showing their performance with an underwater robot in a target following task. Real experiments demonstrate the convergence and stability of the learning system, pointing out its suitability for online robot learning. Advantages and limitations are discussed