The relative performance of ensemble methods with deep convolutional neural networks for image classification

Artificial neural networks have been successfully applied to a variety of machine learning tasks, including image recognition, semantic segmentation, and machine translation. However, few studies fully investigated ensembles of artificial neural networks. In this work, we investigated multiple widely used ensemble methods, including unweighted averaging, majority voting, the Bayes Optimal Classifier, and the (discrete) Super Learner, for image recognition tasks, with deep neural networks as candidate algorithms. We designed several experiments, with the candidate algorithms being the same network structure with different model checkpoints within a single training process, networks with same structure but trained multiple times stochastically, and networks with different structure. In addition, we further studied the overconfidence phenomenon of the neural networks, as well as its impact on the ensemble methods. Across all of our experiments, the Super Learner achieved best performance among all the ensemble methods in this study.     

School-level policies and the efficiency and equity trade-off in education

This paper identifies the relationship between pupils’ Family Background, their mathematics scores, and school-level policies, using the 2012 Programme of International Student Assessment for Italy and multilevel modelling. School-level policies have played a leading role in recent school reforms in many countries, but there is no straightforward empirical evidence for how they affect pupils’ outcomes and the equality of educational outcomes. Our findings show that that only some school policies intensify the Family Background Effect – (educational equity) and affect student outcomes (educational efficiency) simultaneously. We find that competitive schools are able to screen students by selecting higher socio-economic status parents, which mainly harms equity without having much effect on efficiency. There are some policies which allow some trade-off between aforementioned objectives, mainly policies related to management schools.     

Domination and total domination in complementary prisms

Let G be a graph and be the complement of G. The complementary prism of G is the graph formed from the disjoint union of G and by adding the edges of a perfect matching between the corresponding vertices of G and . For example, if G is a 5-cycle, then is the Petersen graph. In this paper we consider domination and total domination numbers of complementary prisms. For any graph G, and , where γ(G) and γ _t (G) denote the domination and total domination numbers of G, respectively. Among other results, we characterize the graphs G attaining these lower bounds. Research supported in part by the South African National Research Foundation and the University of KwaZulu-Natal.     

Beyond the Certainty of Security: On Dominant and Latent Conceptions of “Security” in Modern Societies

While other political programs have a relatively short history, the programatic logic of “security” is rather long-standing. This article argues that “security as a product” is a conception of modernity which came to its limits. The question arises, which kind of conception could he beyond those limits.     

Enhanced rate of resting energy expenditure in women using hormone-replacement therapy: preliminary results

Hormone-replacement therapy (HRT) attenuates the menopause-associated alterations in body composition. It is not known, however, whether this effect is a result of a concomitant increase in energy expenditure. The authors examined whether women submitted to a long-term HRT treatment presented greater energy expenditure than women who had never used HRT. We compared 13 postmenopausal women using HRT (>1 yr) with 13 age- (+/-2 yr) and body-mass-index-matched (BMI; +/-1.5 kg/m2) postmenopausal women not using HRT. Resting energy expenditure (REE; indirect calorimetry), body composition, and daily (DEE) and physical activity (PAEE) energy expenditure (accelerometry) were obtained. Although BMI, fat mass, fat-free mass, DEE, and PAEE were similar between groups, the HRT group displayed a significantly greater REE than the no-HRT group (Delta +222 kcal/day). In conclusion, the authors observed that a long-term treatment with HRT is associated with a greater REE in postmenopausal women. These results need to be confirmed.     

A demographic analysis of the family structure experiences of children in the United States

This paper analyzes the family structure experiences of children in the U.S. Childbearing and transitions among single, cohabiting, and married states are analyzed jointly. A novel contribution is to distinguish men by their relationship to children: biological father or stepfather. The analysis uses data from the NLSY79. A key finding is that children of black mothers spend on average only 33% of their childhood living with the biological father and mother, compared to 74% for children of white mothers. The two most important proximate demographic determinants of the large racial gap are the much higher propensity of black women to conceive children outside of a union, and the lower rate of “shotgun” unions for blacks compared to whites. Another notable finding is that cohabitation plays a negligible role in the family structure experiences of children of white mothers, and even for children of black mothers accounts for less than one fifth of time spent living with both biological parents.

Olympic medals as an indicator of social welfare

The relation between the amount of medals won at the Olympic games and various welfare indicators has been established for the games of Barcelona and Seoul using a Bayesian method of Poisson regression. It was found that medals won correlate strongly with income as well as with more general welfare indicators and that the elasticity with respect to the size of the population is surprisingly smaller than unity.     

BOOK TRANSLATIONS AS IDEA FLOWS: THE EFFECTS OF THE COLLAPSE OF COMMUNISM ON THE DIFFUSION OF KNOWLEDGE

We use book translations as a new measure of international idea flows and study the effects of Communism's collapse in Eastern Europe on these flows. Using novel data on 800,000 translations and difference‐in‐differences approaches, we show that while translations between Communist languages decreased by two thirds with the collapse, Western‐to‐Communist translations increased by a factor of 4 and quickly converged to Western levels. Convergence was more pronounced in the fields of applied and social sciences, and was more complete in Satellite and Baltic than in Soviet countries. We discuss how these patterns help us understand how repressive institutions and preferences towards Western European ideas shaped the international diffusion of knowledge.