From data topology to a modular classifier

This article describes an approach to designing a distributed and modular neural classifier. This approach introduces a new hierarchical clustering that enables one to determine reliable regions in the representation space by exploiting supervised information. A multilayer perceptron is then associated with each of these detected clusters and charged with recognizing elements of the associated cluster while rejecting all others. The obtained global classifier is comprised of a set of cooperating neural networks and completed by a K-nearest neighbor classifier charged with treating elements rejected by all the neural networks. Experimental results for the handwritten digit recognition problem and comparison with neural and statistical nonmodular classifiers are given.Received: 1 October 2002, Accepted: 21 November 2002, Published online: 6 June 2003     

Some Physical Properties of CuTlSe Alloy

The structure and optical properties of Cu0.1Tl2Se3 alloy was studied. X-ray analysis showed that the polycrystalline powder consists of a mixture of two phases Cu3TlSe2 (monoclinic) and TlSe (tetragonal). Chalcopyrite phase CuTlSe2 can be obtained by annealing the powder at 250 ℃ for 1 h. Thin film with a monoclinic phase of Cu3TlSe2 was obtained for the thick film (0.7 mum). The optical parameters alpha, n, k, epsilon' and epsilon " and the energy gap were calculated. The thermoelectric power measurements showed that S has a positive polarity over the whole range of temperature.     

The influence of baking fuel on residues of polycyclic aromatic hydrocarbons and heavy metals in bread

The influence of fuel type used to bake bread on the spectrum and concentrations of some polycyclic aromatic hydrocarbons and heavy metals in baked bread was assessed. Bread samples were collected from different bakeries operated by either electricity, solar, mazot or solid waste and their residue content of PAHs and heavy metals was assessed. The total concentration of PAHs detected in mazot, solar, solid waste and electricity operated bakeries had an average of 320.6, 158.4, 317.3 and 25.5 μg kg⁻¹, respectively. Samples collected from mazot, solar and solid waste operated bakeries have had a wide spectrum of PAHs, in comparison to that detected in bread samples collected from electricity operated bakeries. Lead had the highest concentrations in the four group of bread samples, followed by nickel, while the concentrations of zinc and cadmium were the least. The concentration of lead detected in bread samples produced from mazot, solar, solid waste and electricity fueled bakeries were 1375.5, 1114, 1234, and 257.3 μg kg⁻¹, respectively. Estimated daily intake of PAHs based on bread consumption were 48.2, 28.5, 80.1, and 4.8 μg per person per day for bread produced in bakeries using mazot, solar, solid waste and electricity, respectively. Meanwhile, the estimated daily intake of benzo (a) pyrene were 3.69, 2.65, 8.1, and 0.81 μg per person per day for bread sample baked with mazot, solar, solid waste and electricity, respectively. The daily intake of lead, based on bread consumption was 291, 200.5, 222, and 46.31 μg per person per day for bread sample baked with mazot, solar, solid waste and electricity, respectively. The present work has indicated the comparatively high level of daily intake of benzo (a) pyrene and lead in comparison to levels reported from many other countries and those recommended by international regulatory bodies. It is probable that residues detected in bread samples are partially cereal-borne but there is strong evidence that the process of baking and the gases emitted are responsible for most of the contamination load.     

Sentence alignment using P-NNT and GMM

Parallel corpora have become an essential resource for work in multilingual natural language processing. However, sentence aligned parallel corpora are more efficient than non-aligned parallel corpora for cross-language information retrieval and machine translation applications. In this paper, we present two new approaches to align English–Arabic sentences in bilingual parallel corpora based on probabilistic neural network (P-NNT) and Gaussian mixture model (GMM) classifiers. A feature vector is extracted from the text pair under consideration. This vector contains text features such as length, punctuation score, and cognate score values. A set of manually prepared training data was assigned to train the probabilistic neural network and Gaussian mixture model. Another set of data was used for testing. Using the probabilistic neural network and Gaussian mixture model approaches, we could achieve error reduction of 27% and 50%, respectively, over the length based approach when applied on a set of parallel English–Arabic documents. In addition, the results of (P-NNT) and (GMM) outperform the results of the combined model which exploits length, punctuation and cognates in a dynamic framework. The GMM approach outperforms Melamed and Moore’s approaches too. Moreover these new approaches are valid for any languages pair and are quite flexible since the feature vector may contain more, less or different features, such as a lexical matching feature and Hanzi characters in Japanese–Chinese texts, than the ones used in the current research.     

Volume Changes of Iron Oxide Compacts under Isothermal Reduction Conditions

Compacts made from chemically grade Fe₂O₃ were fired at 1473K for 6 hrs. The fired compacts were isothermally reduced either by hydrogen or carbon monoxide at 1073–1373K. The O₂ weight‐loss resulting from the reduction process was continuously recorded as a function of time using TGA technique, whereas the volume change at different reduction conditions was measured by displacement method. Porosity measurements, microscopic examination and X‐ray diffraction analysis were used to characterize the fired and reduced products. The rate of reduction at both the initial and final stages was increased with temperature. The reduction mechanism deduced from the correlations between apparent activation energy values, structure of partially reduced compacts and application of gas‐solid reaction models revealed the reduction rate (dr/dt) at both the initial and final stages. At early stages, the reduction was controlled by a combined effect of gaseous diffusion and interfacial chemical reaction mechanism, while at the final stages the interfacial chemical reaction was the rate determining step. In H₂ reduction, maximum swelling (80%) was obtained at 1373K, which was attributed to the formation of metallic iron plates. In CO reduction, catastrophic swelling (255%) was obtained at 1198K due to the formation of metallic iron plates and whiskers.     

Preparation and characterization of ZnSe nanoparticles by mechanochemical process

Nanocrystalline ZnSe powder was prepared by mechanochemical process from Zn and Se pure granules. Zn and Se granules transformed to powder during mechanochemical reaction and ZnSe phase was observed after milling time of 16 min by XRD measurements. The crystallite size was measured to be 9–10 nm, while the powder particle size was found to be in the range 90–500 nm in the milling time range 20–60 min as measured by size analyzer and SEM. The homogeneity of the powder was confirmed by EDX analysis. The optical band gap was estimated from optical absorbance curve and possess the value in the range 2.58–2.95 eV. The absorbance peak was observed to be wider according the wide distribution of the crystallite size which affect on the energy gap according to energy gap confinement.     

Flexural and Interfacial Behavior of FRP-Strengthened Reinforced Concrete Beams

Although there is a large amount of experimental data available on the fiber-reinforced polymer (FRP) strengthening of concrete structures, a full understanding of the various debonding phenomena is somewhat lacking. As a contribution to fill this need, two-dimensional and three-dimensional (3D) nonlinear displacement-controlled finite-element (FE) models are developed to investigate the flexural and FRP/concrete interfacial responses of FRP-strengthened reinforced concrete beams. Interface elements are used to simulate the FRP/concrete interfacial behavior before and after cracking. The analysis is carried out using two different relations for the interface; namely, nonlinear and bilinear bond–slip laws. The results predicted using these two laws are compared to those based on the full-bond assumption. The FE models are capable of simulating the various failure modes, including debonding of the FRP, either at the plate end or at intermediate cracks. The 3D model is created to accommodate cases of FRP-strengthened reinforced concrete beams utilizing FRP anchorage systems. In addition, the models successfully represent the actual interfacial behavior at the vicinities of cracks including the stress/slip concentrations and fluctuations. Results are presented in terms of the ultimate load carrying capacities, failure modes and deformational characteristics. Special emphasis is placed on the FRP/concrete interfacial behavior and cracking of the concrete. The numerical results are compared to available experimental data for 25 specimens categorized in six series, and they show a very good agreement.     

Successful Delivery of Public-Private Partnerships for Infrastructure Development

Two common approaches have been used by governments for the implementation of public-private partnerships (PPPs): a finance-based approach that aims to use private financing to satisfy infrastructure needs, and a service-based approach that aims to optimize the time and cost efficiencies in service delivery. The implementation of PPPs, however, may suffer from legal, political, and cultural impediments. In the United States, the federal government enabled a number of acts to ease the impediments and promote PPPs for infrastructure development. Based on a detailed analysis of PPPs in the United Kingdom and British Columbia, Canada, this paper describes principles that would characterize the implementation of PPPs at the program level (e.g., whether the implementation is successful). The principles pertain to the: availability of a PPP legal framework and implementation units; perception of the private finance objectives, risk allocation consequences, and value-for-money objectives; maintenance of PPPs process transparency; standardization of procedures; and use of performance specifications. Guidelines for successful implementation are explained and discussed in the context of the United States PPPs experience and impediments.     

VANET QoS-OLSR: QoS-based clustering protocol for Vehicular Ad hoc Networks

In this paper, we address the problem of clustering in Vehicular Ad hoc Networks (VANETs) using Quality of Service Optimized Link State Routing (QoS-OLSR) protocol. Several clustering algorithms have been proposed for VANET and MANET. However, the mobility-based algorithms ignore the Quality of Service requirements that are important for VANET safety, emergency, and multimedia services while the QoS-based algorithms ignore the high speed mobility constraints since they are dedicated for Mobile Ad hoc Networks (MANETs). Our solution is a new QoS-based clustering algorithm that considers a tradeoff between QoS requirements and high speed mobility constraints. The goal is to form stable clusters and maintain the stability during communications and link failures while satisfying the Quality of Service requirements. This is achieved by: (1) considering the high mobility metrics while computing the QoS, (2) using Ant Colony Optimization for MPRs selection, and (3) using MPR recovery algorithm able to select alternatives and keep the network connected in case of link failures. Performance analysis and simulation results show that the proposed model can maintain the network stability, reduce the end-to-end delay, increase the packet delivery ratio, and reduce the communications overhead.     

Electrocatalytic evolution of hydrogen on a novel SrPdO3 perovskite electrode

SrPdO₃ was prepared for the first time by the citrate method. XRD, SEM and TGA characterizations were carried out. The catalytic activity toward hydrogen evolution reaction (HER) was investigated, the activation energy, and reaction order and reaction mechanism have been determined using Tafel polarization and impedance techniques. The modified surface showed up to 100 times more efficiency towards electrocatalytic production of hydrogen. Adsorption of hydrogen on the catalyst was the rate-determining step and the reaction order at the surface of the catalyst is 0.86. The molar magnetic susceptibility was measured using Faraday's method and anti-ferromagnetic character was observed.