Particle competition and cooperation to prevent error propagation from mislabeled data in semi-supervised learning

Semi-supervised learning is applied to classification problems where only a small portion of the data items is labeled. In these cases, the reliability of the labels is a crucial factor, because mislabeled items may propagate wrong labels to a large portion or even the entire data set. This paper aims to address this problem by presenting a graph-based (network-based) semi-supervised learning method, specifically designed to handle data sets with mislabeled samples. The method uses teams of walking particles, with competitive and cooperative behavior, for label propagation in the network constructed from the input data set. The proposed model is nature-inspired and it incorporates some features to make it robust to a considerable amount of mislabeled data items. Computer simulations show the performance of the method in the presence of different percentage of mislabeled data, in networks of different sizes and average node degree. Importantly, these simulations reveals the existence of the critical points of the mislabeled subset size, below which the network is free of wrong label contamination, but above which the mislabeled samples start to propagate their labels to the rest of the network. Moreover, numerical comparisons have been made among the proposed method and other representative graph-based semi-supervised learning methods using both artificial and real-world data sets. Interestingly, the proposed method has increasing better performance than the others as the percentage of mislabeled samples is getting larger. © 2012 IEEE.

Error estimates for a neumann problem in highly oscillating thin domains

In this work we analyze the convergence of solutions of the Poisson equation with Neumann boundary conditions in a two-dimensional thin domain with highly oscillatory behavior. We consider the case where the height of the domain, amplitude and period of the oscillations are all of the same order, and given by a small parameter e > 0. Using an appropriate corrector approach, we show strong convergence and give error estimates when we replace the original solutions by the first-order expansion through the Multiple-Scale Method.

Sequential Sampling Plan for Tenuipalpus heveae Baker (Acari: Tenuipalpidae) on Rubber Tree

The objective of the present study was to develop a sequential sampling plan for the decision-making process to control Tenuipalpus heveae Baker (Acari: Tenuipalpidae), an important pest of the rubber tree crop. The experimental area was represented by 1,000 plants of the RRIM 600 clone divided in 100 plots with 10 plants each. Leaves were collected and the number of mites determined under laboratory conditions. The sequential sampling plan was developed in accordance with the Sequential Test Likelihood Ratio. The value 0.10 was pre-established for α and β representing type I and type II errors, respectively. The level of control adopted was six mites per 12 cm2. The operating characteristic curve and the curve of maximum expected sample were determined. Two lines were generated: the upper one, when the condition for chemical control is recommended (S1 = 23.3080 + 2.1972); and the lower, when chemical control is not recommended (S0 = -23.3080 + 2.1972). Sample size for the decision-making process to control T. heveae requires 6 to 18 plants. © 2013 Sociedade Entomológica do Brasil.

Global sampling reveals low genetic diversity within Compsopogon (Compsopogonales, Rhodophyta)

Twenty-five specimens of the freshwater red alga Compsopogon were collected from locations in North America, South America, Europe, Asia, Australasia and Oceania, and from an aquarium, with the goal of determining genetic diversity among specimens and ascertaining the number of phylogenetic species. Specimens were morphologically identified as having either the 'caeruleus' morphology, with regular polyhedral cortical cells, or the 'leptoclados' morphology, with irregular cortical cells with rhizoidal outgrowths. The 'leptoclados' morphology has been used by some researchers to distinguish the genus Compsopogonopsis from Compsopogon, or at least to distinguish C. leptoclados from other Compsopogon species. Sequence data for the rbcL gene and cox1 barcoding region were obtained for most specimens. In addition, SSU and partial LSU (barcode) rDNA were explored for a few specimens, but all sequences were identical. For the 25 newly generated and eight previously published rbcL gene data, there were seven unique haplotypes, but the sequence divergence was very low (≤7 bp, ≤ 0.7%). One haplotype was widespread, represented by 21 specimens from diverse locations in all regions sampled. Likewise, the 22 new and one previously published cox1 barcode region sequences yielded seven unique haplotypes with little sequence divergence (≤13 bp, ≤ 2.0%). One haplotype was widespread, being shared among 16 specimens from all regions. The combined molecular and morphological data showed no genetic differentiation between the 'caeruleus' and 'leptoclados' morphologies. The ubiquitous distribution of Compsopogon in tropical/subtropical regions and its low genetic variation are probably facilitated by the alga's ability to tolerate a wide range of stream conditions and its propagation via asexual spores. Given the findings of previous culture-based studies, morphometric research and field observations, coupled with the results of our study, we conclude there is only a single monospecific genus worldwide and that the species is correctly called C. caeruleus, since this is the oldest validly published name; all other previously described species of Compsopogon and Compsopogonopsis are synonyms. © 2013 British Phycological Society.

Influence of visual feedback sampling on obstacle crossing behavior in people with Parkinson's disease

The purpose of the current study was to investigate the role of visual information on gait control in people with Parkinson's disease as they crossed over obstacles. Twelve healthy individuals, and 12 patients with mild to moderate Parkinson's disease, walked at their preferred speeds along a walkway and stepped over obstacles of varying heights (ankle height or half-knee height), under three visual sampling conditions: dynamic (normal lighting), static (static visual samples, similar to stroboscopic lighting), and voluntary visual sampling. Subjects wore liquid crystal glasses for visual manipulation. In the static visual sampling condition only, the patients with Parkinson's disease made contact with the obstacle more often than did the control subjects. In the successful trials, the patients increased their crossing step width in the static visual sampling condition as compared to the dynamic and voluntary visual sampling conditions; the control group maintained the same step width for all visual sampling conditions. The patients showed lower horizontal mean velocity values during obstacle crossing than did the controls. The patients with Parkinson's disease were more dependent on optic flow information for successful task and postural stability than were the control subjects. Bradykinesia influenced obstacle crossing in the patients with Parkinson's disease. © 2013 Elsevier B.V.

A method for improving the code rate and error correction capability of a cyclic code

Currently, there has been an increasing demand for operational and trustworthy digital data transmission and storage systems. This demand has been augmented by the appearance of large-scale, high-speed data networks for the exchange, processing and storage of digital information in the different spheres. In this paper, we explore a way to achieve this goal. For given positive integers n,r, we establish that corresponding to a binary cyclic code C0[n,n-r], there is a binary cyclic code C[(n+1)3k-1,(n+1)3k-1-3kr], where k is a nonnegative integer, which plays a role in enhancing code rate and error correction capability. In the given scheme, the new code C is in fact responsible to carry data transmitted by C0. Consequently, a codeword of the code C0 can be encoded by the generator matrix of C and therefore this arrangement for transferring data offers a safe and swift mode. © 2013 SBMAC - Sociedade Brasileira de Matemática Aplicada e Computacional.

Evaluation of three sampling methods for the microbiological analysis of broiler carcasses after immersion chilling

Countries have different official programs and implement different sampling methods for the detection of Salmonella on poultry carcasses. In Brazil, a 25-g sample of skin and muscle excision (SME) from the wings, neck, and pericloacal parts is used; in the European Union (EU), a 25-g sample of neck skin (NSE) is used; and, in the United States, the whole carcass is rinsed with 400 ml of diluent (WCR). In the present study, these methods were evaluated to compare Salmonella occurrence and counts of hygiene indicator microorganisms (Escherichia coli, Enterobacteriaceae, and total viable count of aerobic mesophilic bacteria) using different carcasses from the same flock and also using different analytical units taken from the same carcass. Eighty flocks, with four broiler carcasses from each, were included in this study; three broilers were sampled according to protocols from Brazil, the EU, and the United States, and the last one by all three methods. SME, NSE, and WCR provided equivalent results (P > 0.05) for Salmonella detection on broiler carcasses when using different carcasses from the same flock and when using the same carcass. The predominant serovar was Salmonella Enteritidis. For the enumeration of hygiene indicator microorganisms, WRC provided higher counts than SME or NSE (P < 0.05), when using both the same or different carcasses. Therefore, it is possible to directly compare Salmonella results in poultry carcasses when using the methods recommended by the legislative bodies of Brazil, the United States, and the EU. However, WCR provides the best results for hygiene indicator microorganisms. Copyright © International Association for Food Protection.

Report of the Workshop on Development and Application of Sampling Frames and Sample Designs for Integrated Household Survey Programmes in the Caribbean

Includes bibliography

Planejamento econômico de gráficos de controle X para monitoramento de processos autocorrelacionados

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Metodologia de amostragens para monitoramento do ácaro da leprose Brevipalpus phoenicis (Geijskes, 1939) em cultura de citros

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)