Particle Competition and Cooperation in Networks for Semi-Supervised Learning

Semi-supervised learning is one of the important topics in machine learning, concerning with pattern classification where only a small subset of data is labeled. In this paper, a new network-based (or graph-based) semi-supervised classification model is proposed. It employs a combined random-greedy walk of particles, with competition and cooperation mechanisms, to propagate class labels to the whole network. Due to the competition mechanism, the proposed model has a local label spreading fashion, i.e., each particle only visits a portion of nodes potentially belonging to it, while it is not allowed to visit those nodes definitely occupied by particles of other classes. In this way, a "divide-and-conquer" effect is naturally embedded in the model. As a result, the proposed model can achieve a good classification rate while exhibiting low computational complexity order in comparison to other network-based semi-supervised algorithms. Computer simulations carried out for synthetic and real-world data sets provide a numeric quantification of the performance of the method.

Does group size matter? Cheating and cooperation in Brazilian school children

Cooperation between individuals is an important requisite for the maintenance of social relationships. The purpose of this study was to investigate cooperation in children in the school environment, where individuals could cooperate or not with their classmates in a public goods game. We investigated which of the following variables influenced cooperation in children: sex, group size, and information on the number of sessions. Group size was the only factor to significantly affect cooperation, with small-group children cooperating significantly more than those in large groups. Both sex and information had no effect on cooperation. We suggest that these results reflect the fact that, in small groups, individuals were more efficient in controlling and retaliating theirs peers than in large groups. (C) 2008 Elsevier Inc. All rights reserved.

The relationship between banking market competition and risk-taking: Do size and capitalization matter?

This paper addresses the effects of bank competition on the risk-taking behaviors of banks in 10 Latin American countries between 2003 and 2008. We conduct our empirical approach in two steps. First, we estimate the Boone indicator, which is a measure of competition. We then regress this measure and other explanatory variables on the banking "stability inefficiency" derived simultaneously from the estimation of a stability stochastic frontier. Unlike previous findings, this paper concludes that competition affects risk-taking behavior in a non-linear way as both high and low competition levels enhance financial stability, while we find the opposite effect for average competition. In addition, bank size and capitalization are essential factors in explaining this relationship. On the one hand, the larger a bank is, the more it benefits from competition. On the other hand, a greater capital ratio is advantageous for banks that operate in collusive markets, while capitalization only enhances the stability of larger banks under high and average competition. These results are of extreme importance when considering bank regulations, especially in light of the recent turmoil in the global financial markets. (C) 2012 Elsevier B.V. All rights reserved.

Effect of the relationship between particle size, inter-particle distance, and metal loading of carbon supported fuel cell catalysts on their catalytic activity

The effect of the relationship between particle size (d), inter-particle distance (x(i)), and metal loading (y) of carbon supported fuel cell Pt or PtRu catalysts on their catalytic activity, based on the optimum d (2.5-3 nm) and x(i)/d (>5) values, was evaluated. It was found that for y < 30 wt%, the optimum values of both d and x(i)/d can be always obtained. For y >= 30 wt%, instead, the positive effect of a thinner catalyst layer of the fuel cell electrode than that using catalysts with y < 30 wt% is concomitant to a decrease of the effective catalyst surface area due to an increase of d and/or a decrease of x(i)/d compared to their optimum values, with in turns gives rise to a decrease in the catalytic activity. The effect of the x(i)/d ratio has been successfully verified by experimental results on ethanol oxidation on PtRu/C catalysts with same particle size and same degree of alloying but different metal loading. Tests in direct ethanol fuel cells showed that, compared to 20 wt% PtRu/C, the negative effect of the lower x(i)/d on the catalytic activity of 30 and 40 wt% PtRu/C catalysts was superior to the positive effect of the thinner catalyst layer.

Effects of the particle sizes and concentrations on the X-ray absorption by CuO compounds

This work presents a study on the effects of the particle size, material concentration and radiation energy on the X-ray absorption. CuO nanoparticles and microparticles were incorporated separately into a polymeric resin in concentrations of 5%, 10% and 30% relative to the resin mass. X-ray absorption by these materials was analyzed with a CdTe detector. The X-ray absorption is higher for the nanostructured material compared to the microstructured one for low energy X-ray beams for all CuO concentrations. (c) 2011 Elsevier Ltd. All rights reserved.

Influence of Particle Size, Applied Compression, and Substratum Material on Particle-Surface Adhesion Force Using the Centrifuge Technique

The centrifuge technique was used to investigate the influence of particle size, applied compression, and substrate material (stainless steel, glass, Teflon, and poly(vinyl chloride)) on particle-surface adhesion force. For this purpose, phosphatic rock (rho(p) = 3090 kg/m(3)) and manioc starch particles (rho(p) = 1480 kg/m(3)) were used as test particles. A microcentrifuge that reached a maximum rotation speed of 14 000 rpm and which contained specially designed centrifuge tubes was used in the adhesion force measurements. The curves showed that the adhesion force profile followed a normal log distribution. The adhesion force increased linearly with particle size and with the increase of each increment of compression force. The manioc starch particles presented greater adhesion forces than the phosphatic rock particles for all particle sizes studied. The glass substrate showed a higher adherence than the other materials, probably due to its smoother topographic surface roughness in relation to the other substrata.

Temporal evolution of tubular initial conditions and their influence on two-particle correlations in relativistic nuclear collisions

Relativistic nuclear collisions data on two-particle correlations exhibit structures as function of relative azimuthal angle and rapidity. A unified description of these near-side and away-side structures is proposed for low to moderate transverse momentum. It is based on the combined effect of tubular initial conditions and hydrodynamical expansion. Contrary to expectations, the hydrodynamics solution shows that the high-energy density tubes (leftover from the initial particle interactions) give rise to particle emission in two directions and this is what leads to the various structures. This description is sensitive to some of the initial tube parameters and may provide a probe of the strong interaction. This explanation is compared with an alternative one where some triangularity in the initial conditions is assumed. A possible experimental test is suggested. (C) 2012 Elsevier B.V. All rights reserved.