Stable feature selection with support vector machines

The support vector machine (SVM) is a popular method for classification, well known for finding the maximum-margin hyperplane. Combining SVM with l1-norm penalty further enables it to simultaneously perform feature selection and margin maximization within a single framework. However, l1-norm SVM shows instability in selecting features in presence of correlated features. We propose a new method to increase the stability of l1-norm SVM by encouraging similarities between feature weights based on feature correlations, which is captured via a feature covariance matrix. Our proposed method can capture both positive and negative correlations between features. We formulate the model as a convex optimization problem and propose a solution based on alternating minimization. Using both synthetic and real-world datasets, we show that our model achieves better stability and classification accuracy compared to several state-of-the-art regularized classification methods.

Learning vector representation of medical objects via EMR-driven nonnegative restricted Boltzmann machines (eNRBM)

Electronic medical record (EMR) offers promises for novel analytics. However, manual feature engineering from EMR is labor intensive because EMR is complex - it contains temporal, mixed-type and multimodal data packed in irregular episodes. We present a computational framework to harness EMR with minimal human supervision via restricted Boltzmann machine (RBM). The framework derives a new representation of medical objects by embedding them in a low-dimensional vector space. This new representation facilitates algebraic and statistical manipulations such as projection onto 2D plane (thereby offering intuitive visualization), object grouping (hence enabling automated phenotyping), and risk stratification. To enhance model interpretability, we introduced two constraints into model parameters: (a) nonnegative coefficients, and (b) structural smoothness. These result in a novel model called eNRBM (EMR-driven nonnegative RBM). We demonstrate the capability of the eNRBM on a cohort of 7578 mental health patients under suicide risk assessment. The derived representation not only shows clinically meaningful feature grouping but also facilitates short-term risk stratification. The F-scores, 0.21 for moderate-risk and 0.36 for high-risk, are significantly higher than those obtained by clinicians and competitive with the results obtained by support vector machines.

Fast one-class support vector machine for novelty detection

Novelty detection arises as an important learning task in several applications. Kernel-based approach to novelty detection has been widely used due to its theoretical rigor and elegance of geometric interpretation. However, computational complexity is a major obstacle in this approach. In this paper, leveraging on the cutting-plane framework with the well-known One-Class Support Vector Machine, we present a new solution that can scale up seamlessly with data. The first solution is exact and linear when viewed through the cutting-plane; the second employed a sampling strategy that remarkably has a constant computational complexity defined relatively to the probability of approximation accuracy. Several datasets are benchmarked to demonstrate the credibility of our framework.

BacHbpred: support vector machine methods for the prediction of bacterial hemoglobin-like proteins

The recent upsurge in microbial genome data has revealed that hemoglobin-like (HbL) proteins may be widely distributed among bacteria and that some organisms may carry more than one HbL encoding gene. However, the discovery of HbL proteins has been limited to a small number of bacteria only. This study describes the prediction of HbL proteins and their domain classification using a machine learning approach. Support vector machine (SVM) models were developed for predicting HbL proteins based upon amino acid composition (AC), dipeptide composition (DC), hybrid method (AC + DC), and position specific scoring matrix (PSSM). In addition, we introduce for the first time a new prediction method based on max to min amino acid residue (MM) profiles. The average accuracy, standard deviation (SD), false positive rate (FPR), confusion matrix, and receiver operating characteristic (ROC) were analyzed. We also compared the performance of our proposed models in homology detection databases. The performance of the different approaches was estimated using fivefold cross-validation techniques. Prediction accuracy was further investigated through confusion matrix and ROC curve analysis. All experimental results indicate that the proposed BacHbpred can be a perspective predictor for determination of HbL related proteins. BacHbpred, a web tool, has been developed for HbL prediction.

Nutritional intervention to prevent weight gain in patients commenced on olanzapine: a randomized controlled trial

Objective: Olanzapine is the most commonly prescribed atypical antipsychotic medication in Australia. Research reports an average weight gain of between 4.5 and 7 kg in the 3 months following its commencement. Trying to minimize this weight gain in a population with an already high prevalence of obesity, mortality and morbidity is of clinical and social importance. This randomized controlled trial investigated the impact of individual nutrition education provided by a dietitian on weight gain in the 3 and 6 months following the commencement of olanzapine.


Method: Fifty-one individuals (29 females, 22 males) who had started on olanzapine in the previous 3 months (mean length of 27 days ± 20) were recruited through Peninsula Health Psychiatric Services and were randomly assigned to either the intervention (n = 29) or the control group (n = 22). Individuals in the intervention group received six 1 hour nutrition education sessions over a 3-month period. Weight, waist circumference, body mass index (BMI) and qualitative measures of exercise levels, quality of life, health and body image were collected at baseline at 3 and 6 months.


Results: After 3 months, the control group had gained significantly more weight than the treatment group (6.0 kg vs 2.0 kg, p ≤ 0.002). Weight gain of more than 7% of initial weight occurred in 64% of the control group compared to 13% of the treatment group. The control group's BMI increased significantly more than the treatment group's (2 kg/m2vs 0.7 kg/m2, p ≤ 0.03). The treatment group reported significantly greater improvements in moderate exercise levels, quality of life, health and body image compared to the controls. At 6 months, the control group continued to show significantly more weight gain since baseline than the treatment group (9.9 kg vs 2.0 kg, p ≤ 0.013) and consequently had significantly greater increases in BMI (3.2 kg/m2vs 0.8 kg/m2, p ≤ 0.017).

Verification Of The Standard Vector Data Model For Interoperability Of River-Geospatial Information

Recently, two international standard organizations, ISO and OGC, have done the work of standardization for GIS. Current standardization work for providing interoperability among GIS DB focuses on the design of open interfaces. But, this work has not considered procedures and methods for designing river geospatial data. Eventually, river geospatial data has its own model. When we share the data by open interface among heterogeneous GIS DB, differences between models result in the loss of information. In this study a plan was suggested both to respond to these changes in the information envirnment and to provide a future Smart River-based river information service by understanding the current state of river geospatial data model, improving, redesigning the database. Therefore, primary and foreign key, which can distinguish attribute information and entity linkages, were redefined to increase the usability. Database construction of attribute information and entity relationship diagram have been newly redefined to redesign linkages among tables from the perspective of a river standard database. In addition, this study was undertaken to expand the current supplier-oriented operating system to a demand-oriented operating system by establishing an efficient management of river-related information and a utilization system, capable of adapting to the changes of a river management paradigm.

Immigration as a strategic vector for the socio-economic and institutional development of Brazil

Brazil has demonstrated resilience in relation to the recent economic crises and has an auspicious development potential projected for the coming decades, which, linked to the globalization process, provides important opportunities for our people. Gradually we have established ourselves as one of the leading nations in the world and we have become a reference in questions linked to economic equilibrium, development, energy, agriculture and the environment. This international recognition favors the exchange of experiences with other cultures, governments and organizations, bringing with it the possibility of stimulating a dynamic process of development and innovation.

Transformation of brain drain of brain gain: incentives to attract talented workforce

The key for the future of any country, firm or group lies in the talent, skills, experience, knowledge and capabilities of its people. Migration of human capital resource on an international level depicts the impact on the developing country having its highly educated individuals migrating to developed countries known as “Brain Drain.” Therefore, evaluation of short-term and long-term talent needs and impacts on any country is critical. This paper aims to complement the existing theoretical brain drain and brain gain literature, focusing on the interaction between investment in education, training, healthcare and government to attract highly talented individuals to a developing a country. The migration study is inclusive of the analysis of the highly talented resources that have committed to or are planning to resettle in their developing native countries after investing in themselves through education. The motivational factors of these highly talented individuals are evaluated to determine key needs and drives attracting these individuals back to China from a developed country (aka. reserve migration).

Does mixed frequency vector error correction model add relevant information to exchange misalignment calculus? Evidence for United States

Real exchange rate is an important macroeconomic price in the economy and a ects economic activity, interest rates, domestic prices, trade and investiments ows among other variables. Methodologies have been developed in empirical exchange rate misalignment studies to evaluate whether a real e ective exchange is overvalued or undervalued. There is a vast body of literature on the determinants of long-term real exchange rates and on empirical strategies to implement the equilibrium norms obtained from theoretical models. This study seeks to contribute to this literature by showing that it is possible to calculate the misalignment from a mixed ointegrated vector error correction framework. An empirical exercise using United States' real exchange rate data is performed. The results suggest that the model with mixed frequency data is preferred to the models with same frequency variables

Estimates of genetic parameters for visual scores and daily weight gain in Brangus animals

(Co) variance components were estimated for visual scores of conformation (CY), early finishing (PY) and muscling (MY) at 550 days of age (yearling), average daily gain from weaning to yearling (GWY), conformation (CW), early finishing (PW) and muscling (MW) scores at weaning, and average daily gain from birth to weaning (GBW) in animals forming the Brazilian Brangus breed born between 1986 and 2002 from the livestock files of GenSys Consultants Associados S/C Ltda. The data set contained 53 683; 45 136; 52 937; 56 471; 24 531; 21 166; 24 006 and 25 419 records for CW, PW, MW, GBW, CY, PY, MY and GWY, respectively. Data were analyzed by the restricted maximum likelihood method using single-and two-trait animal models. Direct heritability estimates obtained by single-trait analysis were 0.12, 0.14, 0.13 and 0.14 for CY, PY and MY scores and GWY, respectively. A positive association was observed between the same visual scores at weaning and yearling, with correlations ranging from 0.64 to 0.94. Estimated correlations between GBW and weaning and yearling scores ranged from 0.60 to 0.77. The genetic correlation between GBW and GWY was low (0.10), whereas correlations of 0.55, 0.37 and 0.47 were observed between GWY and CY, PY and MY, respectively. Moreover, GWY showed a weak correlation with CW (0.10), PW (-0.08) and MW (-0.03) scores. These results indicate that selection of the traits that was studied would result in a small response. In addition, selection based on average daily gain may have an indirect effect on visual scores as the correlations between GWY and visual scores were generally strong.

Suckling and allosuckling in river buffalo calves and its relation with weight gain

The aim of the present study was to investigate if river buffalo calves (Bubalus bubalis) have equal access to all cows and if milk is thus equally available to all of them. We recorded suckling/allosuckling behaviour and weight gain (WG) of 29 calves (14 males and 15 females), with special consideration to their sex, birth order (BO) and age. Cows' nursing behaviour and milli production (MP) were also considered. While males tended to be born earlier than females during this study, this was not the trend in the overall herd records. The cows' MP was not effected by the calves' sex. However, bull-calves presented greater mean WG, and mean times spent in individual filial (IF) and in communal nonfilial (CNF) suckling than heifer-calves, which showed greater communal filial (CF) suckling than the former during the first 4 months of life. The WG was associated with IF for bull-calves (r = 0.680 and 0.765, respectively, for the periods from birth to 4th and 8th months of age), and to CNF for heifer-calves (r = 0.628, for the period from birth to 8th month). Results from multiple regression analysis showed independent effects of each suckling category on the calf WG, and such effects were variable according to the calf's sex. BO was negatively correlated to calves' WG (bull-calves: r(s) = - 0.873 and - 0.799, from birth to 4th and gth months, respectively; heifer-calves: r(s) = - 0.531 from birth to 4th month). Specifically for bull-calves, there was a positive correlation between BO and MP (r(s) = 0.528 and 0.633, from birth to 4th and 8th months of age, respectively). The correlation between BO and IF was negative in both sexes, indicating that calves that were born early had more opportunities to suckle individually from their mothers. For heifer-calves, BO was positively correlated with CF (two periods), and negatively with CNF (from birth to 8th month of age), suggesting that heifer-calves were most often accompanied by other calves during suckling when they were born later. The data taken together indicate that sex and/or BO influenced decisively social interactions during suckling, promoting differential development among the calves. In animal husbandry, if a homogenous WG is desired, these factors have to be taken into consideration. (C) 2000 Elsevier B.V. B.V. All rights reserved.

Prediction equations to estimate the demand of energy and crude protein for maintenance, gain and egg production for laying Japanese quails

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Prediction of intake and average daily gain by different feeding systems for goats

A main purpose of a mathematical nutrition model (a.k.a., feeding systems) is to provide a mathematical approach for determining the amount and composition of the diet necessary for a certain level of animal productive performance. Therefore, feeding systems should be able to predict voluntary feed intake and to partition nutrients into different productive functions and performances. In the last decades, several feeding systems for goats have been developed. The objective of this paper is to compare and evaluate the main goat feeding systems (AFRC, CSIRO, NRC, and SRNS), using data of individual growing goat kids from seven studies conducted in Brazil. The feeding systems were evaluated by regressing the residuals (observed minus predicted) on the predicted values centered on their means. The comparisons showed that these systems differ in their approach for estimating dry matter intake (DMI) and energy requirements for growing goats. The AFRC system was the most accurate for predicting DMI (mean bias = 91 g/d, P < 0.001; linear bias 0.874). The average ADG accounted for a large part of the bias in the prediction of DMI by CSIRO, NRC, and, mainly, AFRC systems. The CSIRO model gave the most accurate predictions of ADG when observed DMI was used as input in the models (mean bias 12 g/d, P < 0.001; linear bias -0.229). while the AFRC was the most accurate when predicted DMI was used (mean bias 8g/d. P > 0.1; linear bias -0.347). (C) 2011 Elsevier B.V. All rights reserved.

Genetic parameters and relationships of heifer pregnancy and age at first calving with weight gain, yearling and mature weight in Nelore cattle

The objectives of the current study were to investigate the additive genetic associations between heifer pregnancy at 16 months of age (HP16) and age at first calving (AFC) with weight gain from birth to weaning (WG), yearling weight (YW) and mature weight (MW), in order to verify the possibility of using the traits measured directly in females as selection criteria for the genetic improvement of sexual precocity in Nelore cattle. (Co)variance components were estimated by Bayesian inference using a linear animal model for AFC, WG, YW and MW and a nonlinear (threshold) animal model for HP16. The posterior means of direct heritability estimates were: 0.45 +/- 0.02; 0.10 +/- 0.01; 023 +/- 0.02; 0.36 +/- 0.01 and 0.39 +/- 0.04, for HP16, AFC, WG, YW and MW, respectively. Maternal heritability estimate for WG was 0.07 +/- 0.01. Genetic correlations estimated between HP16 and WG, YW and MW were 0.19 +/- 0.04; 0.25 +/- 0.06 and 0.14 +/- 0.05, respectively. The genetic correlations of AFC with WG, YW and MW were low to moderate and negative, with values of -0.18 +/- 0.06; -0.22 +/- 0.05 and -0.12 +/- 0.05, respectively. The high heritability estimated for HP16 suggests that this trait seem to be a better selection criterion for females sexual precocity than AFC. Long-term selection for animals that are heavier at young ages tends to improve the heifers sexual precocity evaluated by HP16 or AFC. Predicted breeding values for HP16 can be used to select bulls and it can lead to an improvement in sexual precocity. The inclusion of HP16 in a selection index will result in small or no response for females mature weight. (C) 2011 Elsevier B.V. All rights reserved.