Gecynalda Soares da Silva Gomes

Childhood anemia prevalence and associated factors in Salvador, Bahia, Brazil

A cross-sectional study was conducted in 1996 to estimate the prevalence of anemia in a stratified sample of 603 preschool children and identify factors associated with the disease. Hemoglobin assays were conducted in finger-prick blood samples using a Hemocue hemoglobinometer. Anemia was defined as hemoglobin below 11.0 g/dl. Logistic regression analysis was used to evaluate the potential associations. Observed anemia prevalence was 46.3%. Associated factors were: the 6-12-month and 12-24-month age brackets, the lowest tertiles for iron density and protein content dietary intake, and any degree of deficit in the height-for-age anthropometric parameter. Inadequate physical, sanitary, and environmental conditions in the home were associated with a significantly increased risk of anemia. Anemia constitutes an important health problem in this studys child population. Improvements in living conditions and dietary quality could contribute to a reduction in anemia prevalence.

Comparison of new activation functions in neural network for forecasting financial time series

In artificial neural networks (ANNs), the activation function most used in practice are the logistic sigmoid function and the hyperbolic tangent function. The activation functions used in ANNs have been said to play an important role in the convergence of the learning algorithms. In this paper, we evaluate the use of different activation functions and suggest the use of three new simple functions, complementary log-log, probit and log-log, as activation functions in order to improve the performance of neural networks. Financial time series were used to evaluate the performance of ANNs models using these new activation functions and to compare their performance with some activation functions existing in the literature. This evaluation is performed through two learning algorithms: conjugate gradient backpropagation with Fletcher–Reeves updates and Levenberg–Marquardt.

Duração do aleitamento materno, regime alimentar e fatores associados segundo condições de vida em Salvador, Bahia, Brasil

This cross-sectional study aimed to identify breastfeeding duration, infant feeding regimes, and factors related to living conditions among 811 children under 24 months of age in Salvador, Bahia, Brazil. Data were statistically analyzed by survival analysis, Pearsons chi-square test, and multivariate logistic regression. Median duration of exclusive, predominant, and total breastfeeding was 30.6, 73.0, and 131.5 days, respectively. Exclusive or predominant breastfeeding was discontinued in 83.6% of the subjects. Children with poor living conditions were 2.3 times more likely (95%CI: 1.09-5.01) to receive early supplementary food, whereas the figure for the very poor increased to 2.5 (95%CI: 1.20-5.34). Early exclusive or predominant breastfeeding discontinuation was associated with early pregnancy and poor living conditions of the children and their families. Programs directed towards proper breastfeeding and healthy feeding practices in childhood should consider the social factors associated with early introduction of supplementary foods in this population.

Níveis de hemoglobina, aleitamento materno e regime alimentar no primeiro ano de vida

OBJETIVO: Identificar a relacao entre os niveis de hemoglobina e o consumo de leite materno, alimentos complementares e liquidos nao nutritivos no primeiro ano de vida. METODOS: Estudo transversal envolvendo 553 criancas menores de 12 meses de vida, que frequentavam os servicos publicos de saude. A concentracao de hemoglobina foi avaliada pelo metodo cianometahemoglobina, usando-se o sistema HemoCue. Utilizou-se a tecnica da regressao linear multipla para avaliar as associacoes de interesse. RESULTADOS: Niveis de hemoglobina compativeis com a anemia foram identificados em 62,8% das criancas investigadas, com maior ocorrencia naquelas de seis a 12 meses de idade (72,6%). O aleitamento materno exclusivo nos primeiros seis meses de vida assegurou os mais elevados niveis de hemoglobina. Os demais regimes alimentares declinaram de maneira diferenciada os niveis de hemoglobina, que se tornaram compativeis com a anemia quando o regime de aleitamento artificial foi adotado (p=0,009). O consumo de cha e/ou agua declinou em 0,76 g/dl (p<0,001) os niveis de hemoglobina dos menores de seis meses de idade. Para as criancas de seis a 12 meses, os niveis de hemoglobina variaram significante e positivamente com o consumo de acucar (p=0,017) e feijao (p=0,018) e negativamente com o consumo de fruta (p<0,001). CONCLUSOES: O aleitamento materno exclusivo ate os seis meses de idade e a manutencao do leite materno a partir dessa idade, associado aos alimentos complementares quali e quantitativamente adequados, podem contribuir para o aumento dos niveis da hemoglobina no primeiro ano de vida.

Optimization of the weights and asymmetric activation function family of neural network for time series forecasting

The use of neural network models for time series forecasting has been motivated by experimental results that indicate high capacity for function approximation with good accuracy. Generally, these models use activation functions with fixed parameters. However, it is known that the choice of activation function strongly influences the complexity and neural network performance and that a limited number of activation functions has been used in general. We describe the use of an asymmetric activation functions family with free parameter for neural networks. We prove that the activation functions family defined, satisfies the requirements of the universal approximation theorem We present a methodology for global optimization of the activation functions family with free parameter and the connections between the processing units of the neural network. The main idea is to optimize, simultaneously, the weights and activation function used in a Multilayer Perceptron (MLP), through an approach that combines the advantages of simulated annealing, tabu search and a local learning algorithm. We have chosen two local learning algorithms: the backpropagation with momentum (BPM) and Levenberg-Marquardt (LM). The overall purpose is to improve performance in time series forecasting.

Determinants of mild-to-moderate malnutrition in preschoolers in an urban area of Northeastern Brazil: a hierarchical approach

OBJECTIVE To investigate the determinants of mild-to-moderate malnutrition in preschoolers. DESIGN Cross-sectional study conducted in October and November 1996, with a representative sample of 1740 children less than 5 years old from the city of Salvador, situated in the Brazilian Northeastern region. Socio-economic and dietary data were collected through a structured questionnaire. Anthropometric measures were performed in duplicate and data analysis was based upon the hierarchical model approach. Logistic regression analysis was used to estimate the prevalence ratio and to identify the determinants of mild-to-moderate deficits in weight-for-age and height-for-age Z-scores. RESULTS Family monthly income under US

Hybrid model with dynamic architecture for forecasting time series

67.00 per capita and family headed by a woman were the main basic determinants of mild-to-moderate weight-for-age and height-for-age deficits in the studied children. Household agglomeration, an underlying determinant, was associated with weight-for-age and height-for-age deficits. Among the immediate determinants, age above 6 months and dietary caloric availability in the lowest tertile (<930 kcal day-1) were also associated with weight-for-age deficits. In addition to these, hospitalisation in the 12 months preceding the interview was shown to be a predictor of mild-to-moderate weight-for-age and height-for-age deficits. CONCLUSION Adverse social and economic factors interact with family environmental factors to define food consumption and morbidity patterns that culminate in a high prevalence of mild-to-moderate malnutrition. The strengthening and restructuring of nutrition and healthcare actions, the definition of public policies that improve family income, and the adequate insertion of women in the labour market are possible strategies to reduce mild-to-moderate malnutrition and to sustain the decline already observed in severe malnutrition.

Evidence of software inspection on feature specification for software product lines

Nonlinear artificial neural network models are very attractive for modeling and forecasting time series. The use of such models in these types of applications is motivated by experimental results that show a high capacity of approximation for functions with high accuracy. However, many researchers have used feedforward and/or backpropagation models for time series predictions. In this paper, a model is applied for neural networks with the dynamic architecture proposed by Ghiassi and Saidane (2005), known as the DAN2 model. The results of DAN2 are compared with auto-regressive integrated mobile average (ARIMA) models. As the main result of the paper, we propose a hybrid model with dynamic architecture (HAD) based on combinations of individual forecasts from the DAN2 and ARIMA models with the aim of obtaining more precise forecasts for poorly behaved time series. The results suggest that for this kind of series, the HAD hybrid model outperforms the individual DAN2 and ARIMA models.

Complementary Log-Log and Probit: Activation Functions Implemented in Artificial Neural Networks

In software product lines (SPL), scoping is a phase responsible for capturing, specifying and modeling features, and also their constraints, interactions and variations. The feature specification task, performed in this phase, is usually based on natural language, which may lead to lack of clarity, non-conformities and defects. Consequently, scoping analysts may introduce ambiguity, inconsistency, omissions and non-conformities. In this sense, this paper aims at gathering evidence about the effects of applying an inspection approach to feature specification for SPL. Data from a SPL reengineering project were analyzed in this work and the analysis indicated that the correction activity demanded more effort. Also, Paretos principle showed that incompleteness and ambiguity reported higher non-conformity occurrences. Finally, the Poisson regression analysis showed that sub-domain risk information can be a good indicator for prioritization of sub-domains in the inspection activity.

Evaluating Lehman’s Laws of Software Evolution within Software Product Lines: A Preliminary Empirical Study

The types of activation functions most often used in artificial neural networks are logistic and hyperbolic tangent. Activation functions used in ANN have been said to play an important role in the convergence of the algorithms used. This paper uses sigmoid functions in the processing units of neural networks. Such functions are commonly applied in statistical regression models. The nonlinear functions implemented here are the inverse of complementary log-log and probit link functions. A Monte Carlo framework is presented to evaluate the results of prediction power with these nonlinear functions.