Search

Aminobenzonitrile adsorption on silver nanoparticles (SNPs) was studied with SERS, SEIRA, and DFT. It was found that 4-aminobenzonitrile (4ABN) and 2-aminobenzonitirle (2ABN) could resonantly distribute an amino group charge so that the cyano group could strongly interact with the SNPs in the mono/multilayer. Cyano stretching frequency shifts in SEIRA spectra showed that a 4ABN multilayer interacted strongly with the SNPs when deposited using an alkane solvent with nonpolar bonds. 2ABN could interact with the SNPs even when deposition occurred using a solvent with polar bonds because intramolecular hydrogen-bonding in 2ABN limited solvation. 3-Aminobenzonitrile (3ABN) never interacted strongly with the SNPs in the multilayer because it could not resonantly distribute the amino group charge. 2ABN also formed C–H hydrogen-bonds with n-heptane, which could interact with both the amino and cyano groups of 2ABN. Charge transfer between adsorbed ABN isomers and the SNPs is evidenced in SERS spectra by the prese...

Orientador : Prof. Dr. Eduardo Todt Coorientador : Prof. Dr. Carlos Sanquetta Tese (doutorado) - Universidade Federal do Paraná, Setor de Ciências Exatas, Programa de Pós-Graduação em Informática. Defesa: Curitiba, 25/11/2016 Inclui referências : f. 76-84 Área de concentração : Ciência da computação Resumo: Vive-se em um mundo onde a escassez de recursos naturais leva a um uso cada vez mais racional destes, seja água, recursos minerais, biológicos, hídricos e energéticos. Especificamente quando se trata de madeira, o homem evoluiu muito no manejo de florestas plantadas, aplicando técnicas de cultivo e planejamento de corte para a exploração. Estas técnicas fazem uso da mensuração florestal para medição e estimação de valores dendrométricos importantes, como altura, volume e biomassa. A estimativa de valores dendrométricos é de extrema importância, pois não é viável o abate de toda uma população para que sejam observados os exatos valores. Assim, pesquisadores fazem uso de ferramentas estatísticas em mensuração florestal há anos, com bastante sucesso em suas estimações. Recentemente, com o avanço da área da inteligência artificial, técnicas de aprendizado de máquina têm se mostrado também capazes de competir com os métodos estatísticos de regressão, abrindo assim um leque de opções aos pesquisadores. Inserido neste contexto, o objetivo deste trabalho é aplicar técnicas de aprendizado de máquina para resolução de problemas de mensuração florestal, mostrando também que este se encaixa como um processo de descoberta de conhecimento, inserido na área da Ciência da Computação. Foram realizados experimentos com dados de árvores de Acácia-negra para avaliação de biomassa e relação hipsométrica, Pinus para estimativa de volumes e relação hipsométrica, e com uma amostra de árvores de florestas tropicais de várias regiões do mundo, para estimativa de biomassa. Compararam-se modelos alométricos clássicos com Redes Neurais Artificiais (RNA), Máquinas de Vetores de Suporte (SVM) e Random Forest (RF), e em todos os testes realizados o modelo que obteve melhor correlação é uma técnica de aprendizado de máquina. Para a predição de volumes de Pinus, o melhor modelo foi SVM com correlação de 99,19%. Para estimativa de biomassa da Acácia-negra, SVM obteve a melhor correlação com 98,6%. Para estimativa de biomassa de florestas tropicais, o melhor modelo foi RNA com correlação de 98,06%. Para relação hipsométrica da Acácia-negra, o modelo de SVM obteve a melhor correlação, de 97,73%. Já para relação hipsométrica de Pinus, o modelo de melhor predição foi RNA com correlação de 98,02%. O teste de Friedman mostrou a presença de diferença estatística entre os métodos e, embora o pós-teste de Nemenyi não tenha conseguido evidenciá-la, foi detectada uma tendência a uma separação entre os métodos. Os resultados obtidos mostram fortemente que os modelos de aprendizado de máquina são uma alternativa competitiva frente aos métodos clássicos, superando-os nos experimentos aqui realizados. Palavras chave: Inventário Florestal, Mensuração Florestal, Descoberta de Conhecimento em Bases de Dados, Aprendizado de Máquina, Redes Neurais, Máquinas de Vetores de Suporte, Árvores de Modelos, Florestas Aleatórias. Abstract: We live in a world where scarcity of natural resources leads to an increasing rational use of these resources, like water, minerals, biological or energetical. Specifically wood, man evolved much in the management of planted forests by applying cultivation techniques and cutting planning for exploration. These techniques make use of forest measurement to estimation of important dendrometric values such as height, volume and biomass. The estimation of dendrometric values is extremely important because is not feasible to cut down the whole population to note the exact values. Thus, researchers make use of statistical tools for measuring forest for years with success in their estimates. Recently, with the advancement of artificial intelligence, machine learning techniques have been able to compete with the statistical regression methods, thereby opening a range of options to researchers. Within this context, the objective of this work is to use machine learning techniques to resolution of forest measurement problems, showing that this is a process of knowledge discovery, from the area of Computer Science. Experiments were performed with data of Acacia-negra trees for evaluation of biomass and hypsometric relation, Pinus to estimate volumes and hypsometric relation, and with a sample of tropical forest trees of various regions to estimate biomass. Classical allometric models were compared with Artificial Neural Networks (ANN), Support Vector Machines (SVM) and Random Forest (RF), and in all tests the model that obtained the best correlation was a machine learning technique. For the prediction of Pinus volumes, the best model was SVM with correlation of 99.19 %. To estimate Acacia-negra biomass, SVM showed the best correlation with 98.6 %. To estimate biomass of tropical forest trees, the best model was RNA with a correlation of 98.06 %. For hypsometric relation of Acacia-negra, the SVM model obtained the best correlation, of 97.73 %. As for the hypsometric relation of Pinus, the best prediction model was RNA with a correlation of 98.02The Friedman test showed the presence of statistical difference between the methods and, although the Nemenyi post-hoc test was not able to show it, it was detectecd a tendency towards a separation of methods. The results obtained strongly show that machine learning models are a competitive alternative in comparison to the classical methods, surpassing them in the experiments carried out here. Keywords: Forest Inventory, Forest Measuring, Knowledge Discovery in Databases Machine Learning, Neural Networks, Support Vector Machines, Model Trees, Random Forest.

BACKGROUND: The objective was to evaluate the performance of nucleic acid testing (NAT) in the detection of hepatitis B virus (HBV) infection in hepatitis B surface antigen (HBsAg)-positive blood donations. STUDY DESIGN AND METHODS: A total of 253 HBsAg- and anti-hepatitis B core antigen (HBc)-positive samples (50 hepatitis B e antigen [HBeAg]-positive and 203 anti-HBe–positive) from blood donations collected in France were studied. The samples were investigated with a blood screening assay (Procleix Ultrio, Chiron/Gen-Probe) in minipool (MP; ×8) and in individual-donation (ID) testing. All nonreactive samples were retested once, and nonreactive MP samples were assayed for viral load (VL). RESULTS: All 50 HBeAg-positive samples were reactive in MP-NAT and ID-NAT. Of the 203 anti-HBe–positive donations, 80.3 percent were MP- and ID-reactive, 17.2 percent were MP-nonreactive and ID-reactive, and 2.5 percent were nonreactive in ID-NAT. Overall the sensitivity of ID-NAT was 98 percent versus 84 percent for MP-NAT. After retesting, 16 of the 35 MP-nonreactive and/or ID-reactive donations became MP-reactive and 2 of the ID-nonreactive donations became NAT-reactive. The capacity of Procleix Ultrio to detect HBV DNA was not related to HBsAg subtype, but correlated with the VL: the mean VL in the group of MP-nonreactive samples was 1,420 copies per mL vs. 17,000 copies per mL in the group of 40 MP-reactive samples. CONCLUSION: These results demonstrate that HBV-NAT in ID format is far more effective in detecting viremia in chronic HBsAg carriers than in MP-NAT. The sensitivity of the NAT assay needs to be improved to be considered for replacing the current HBsAg assays, especially when anti-HBc testing is not performed.

This article demonstrates how action research can fill a theoretical gap in the literature on ‘social exclusion’. Although the literature points to the centrality of ‘relationships’ in overcoming exclusion, it offers little theory on the nature of these relationships and how they are formed. This article presents ‘actionable knowledge’ for creating interpersonal relationships that interrupt processes of exclusion. It draws on the findings of an ‘action science’ inquiry process carried out by the staff of an intervention program that helps schools work more effectively with students characterized by chronic failure. The article describes how the concept of ‘social exclusion’ entered the discourse of the inquiry and helped practitioners frame their practice problem and intervention strategy. It then analyzes the program’s specific ‘theory of action’ for forming inclusive relationships. Finally it addresses the validity and the generalizability of action science research aimed at developing a theory of inclusive practice.

BACKGROUND: The purpose of this study was to compare the performances of HCV core antigen (HCV Ag) testing with HCV RNA detection during the preseroconversion period. STUDY DESIGN AND METHODS: Six HCV antibody (HCV Ab)-negative and HCV RNA-positive blood samples from 6 donors and 135 serial samples from 28 patients who had undergone hemodialysis, collected a mean of 90 days before the detection of HCV Ab, were tested by ELISA for the detection of HCV Ag and by PCR to quantify HCV RNA. RESULTS: Five of the six donors were positive for HCV Ag. The donor with a negative HCV Ag test had the lowest viral load. In the hemodialysis patients, the 43 first specimens of the series were HCV RNA negative. Of the 92 specimens that were HCV RNA positive, 81 (88%) were positive for HCV Ag. Among the 74 samples with more than 105 RNA copies, 71 (96%) were HCV Ag positive. Average time from first viremic bleed to first HCV Ag-positive bleed was estimated at 2.0 days and that to first HCV Ab-positive bleed at 50.8 days. CONCLUSION: HCV Ag testing permits the detection of an HCV infection about 1.5 months earlier than the HCV Ab screening tests and an average of only 2 days later than quantitative HCV RNA detection in individual specimens.

BACKGROUND: Exposure to GB virus type C/HGV (GBV-C/HGV) could be determined by detection either of RNA by RT-PCR or of antibodies of the envelope protein E2. STUDY DESIGN AND METHODS: The aim of the study was to determine the proportion of the GBV-C/HGV markers of infection in a blood donor population infected with HCV and to identify GBV-C/HGV routes of transmission that are associated with HCV genotypes and risk factors. RESULTS: Among 306 HCV RNA-positive blood donors, the proportion of GBV-C/HGV RNA-positive donors and anti-E2-positive donors was 19.3 percent (95% CI = 15.0-24.2%) and 42.1 percent (95% CI = 36.6-47.9%), respectively. Exposure to GBV-C/HGV (RNA or anti-E2) was significantly associated with the risk factor of IV drug use. There was a trend toward association with HCV subtypes 1a and 3a, probably because these HCV subtypes are the most frequent in IV drug users. No correlation was observed between ALT elevation and the presence of GBV-C/HGV RNA. CONCLUSION: In persons with HCV infection, IV drug use seems to be a major route of GBV-C/HGV transmission. Precautions taken to avoid HCV infection will probably also decrease GBV-C/HGV transmission.

Over the course of the last decade, there have been several improvements in the performance of Boolean Satisfiability (SAT), Integer Linear Programming (ILP) and Pseudo-Boolean Optimization (PBO) solvers. These improvements have encouraged the applications of SAT, ILP and PBO techniques in modeling complex engineering problems. One such problem is the Virtual Machine Consolidation. The Virtual Machine Consolidation problem consists in placing a set of virtual machines in a set of hardware in a way to increase workload on hardware where they can operate more energy-efficient. This paper proposes an improved PBO formulation of the Virtual Machine Consolidation problem, PBFVMC. The improved formulation and enhancements are built on top of a previous work and a new set of constraints is created and rationalized to work more friendly with current PBO solvers. It is observed that this new formulation goes a step ahead and more problems can now be solved.

Hepatitis C virus (HCV) isolates have been classified into six main genotypes. Genotyping methods, and especially the widely used line probe assay (LiPA), are frequently based on the 5′-untranslated region (5′UTR). However, this region is not appropriate for discriminating HCV strains at the subtype level and for distinguishing many genotype 6 samples from genotype 1. We investigated the capacity of a novel LiPA (Versant HCV Genotype 2.0 assay) based on the simultaneous detection of 5′UTR and Core regions for genotypes 1 and 6 to provide correct HCV genotypes (characterized with a phylogenetic analysis) in a set of HCV strains mainly encountered in Western countries. The improvement was assessed by comparing the results to those obtained with the previous version of the assay. Of the 135 tested samples, 64.7% were concordant for genotype group and subtype with sequencing reference results using the Versant HCV Genotype 2.0 assay versus 37.5% with the previous version. The yield was mainly related to a better characterization of genotype 1, since the accuracy, tested in 62 genotype 1 samples, increased from 45.2% with the first version to 96.8% with the new one. However, this new version necessitates a specific PCR and could no longer be used after 5′UTR PCR used for current HCV infection diagnosis. Moreover, the information provided by 5′UTR hybridization is not reliable for correctly identifying the diversity within genotypes 2 and 4. Thus, the Versant HCV Genotype 2.0 assay remains a useful tool for clinical practice when only the discrimination between major HCV genotypes is necessary.

Orientador : Fabiano Silva Dissertação (mestrado) - Universidade Federal do Paraná, Setor de Ciencias Exatas, Programa de Pós-Graduação em Informática. Defesa: Curitiba, 2006 Inclui bibliografia

National audience