ECG biometrics using a dissimilarity space representation

Electrocardiogram (ECG) biometrics are a relatively recent trend in biometric recognition, with at least 13 years of development in peer-reviewed literature. Most of the proposed biometric techniques perform classifi-cation on features extracted from either heartbeats or from ECG based transformed signals. The best representation is yet to be decided. This paper studies an alternative representation, a dissimilarity space, based on the pairwise dissimilarity between templates and subjects' signals. Additionally, this representation can make use of ECG signals sourced from multiple leads. Configurations of three leads will be tested and contrasted with single-lead experiments. Using the same k-NN classifier the results proved superior to those obtained through a similar algorithm which does not employ a dissimilarity representation. The best Authentication EER went as low as 1:53% for a database employing 503 subjects. However, the employment of extra leads did not prove itself advantageous.

Text classification using compression-based dissimilarity measures

Arguably, the most difficult task in text classification is to choose an appropriate set of features that allows machine learning algorithms to provide accurate classification. Most state-of-the-art techniques for this task involve careful feature engineering and a pre-processing stage, which may be too expensive in the emerging context of massive collections of electronic texts. In this paper, we propose efficient methods for text classification based on information-theoretic dissimilarity measures, which are used to define dissimilarity-based representations. These methods dispense with any feature design or engineering, by mapping texts into a feature space using universal dissimilarity measures; in this space, classical classifiers (e.g. nearest neighbor or support vector machines) can then be used. The reported experimental evaluation of the proposed methods, on sentiment polarity analysis and authorship attribution problems, reveals that it approximates, sometimes even outperforms previous state-of-the-art techniques, despite being much simpler, in the sense that they do not require any text pre-processing or feature engineering.

Similar understorey structure in spite of edaphic and floristic dissimilarity in Amazonian forests

Forest structure determines light availability for understorey plants. The structure of lowland Amazonian forests is known to vary over long edaphic gradients, but whether more subtle edaphic variation also affects forest structure has not beenresolved. In western Amazonia, the majority of non-flooded forests grow on soils derived either from relatively fertile sediments of the Pebas Formation or from poorer sediments of the Nauta Formation. The objective of this study was to compare structure and light availability in the understorey of forests growing on these two geological formations. We measured canopy openness and tree stem densities in three size classes in northeastern Peru in a total of 275 study points in old-growth terra firme forests representing the two geological formations. We also documented variation in floristic composition (ferns, lycophytes and the palm Iriartea deltoidea) and used Landsat TM satellite image information to model the forest structural and floristic features over a larger area. The floristic compositions of forests on the two formations were clearly different, and this could also be modelled with the satellite imagery. In contrast, the field observations of forest structure gave only a weak indication that forests on the Nauta Formation might be denser than those on the Pebas Formation. The modelling of forest structural features with satellite imagery did not support this result. Our results indicate that the structure of forest understorey varies much less than floristic composition does over the studied edaphic difference.

A measure of group dissimilarity for psychological attributes

El funcionamiento y el rendimiento de los grupos en contextos diferentes están relacionados con el grado en que las características de los miembros son complementarias o suplementarias. El presente artículo describe un procedimiento para cuantificar el grado de disimilitud a nivel de grupo. A diferencia de la mayoría de técnicas existentes, el procedimiento que aquí se describe está normalizado y es invariante a los cambios de localización y escala. Por lo tanto, es posible comparar la disimilitud en escalas con diferente métrica y en grupos de distinto tamaño. La disimilitud está medida en términos relativos, independientemente de la posición que ocupan los individuos en la dimensión que mide la escala. Cuando no existe una justificación teórica para combinar las diversas propiedades medidas, se puede cuantificar la disimilitud para cada escala por separado. También es posible obtener las contribuciones diádicas e individuales respecto a la diversidad global y la asignada a cada escala. Las medidas descriptivas pueden ser complementadas con la significación estadística para, así, comparar los resultados obtenidos con distribuciones discretas de referencia, ya sean simétricas o asimétricas. Se ha elaborado un paquete en R que permite obtener los índices descriptivos y los valores p, además de contener las expresiones desarrolladas para simular una amplia variedad de distribuciones discretas de probabilidad.

Genetic dissimilarity among jabuticaba trees native to southwestern Paraná, Brazil

Knowledge on the genetic diversity within and between genotype groups is of great importance for breeding programs. The purpose of this study was to estimate the genetic dissimilarity among 36 native jabuticaba trees (Plinia cauliflora) from five sites in the southwestern region of Paraná, Brazil. Sixteen fruit traits were analyzed, based on multivariate techniques (canonical variables, Tocher and UPGMA), using Mahalanobis' distance as dissimilarity measure. By the techniques of clustering and graphic dispersion, together with the comparison of means, the genetic diversity among native jabuticaba trees was efficiently identified, indicating a high potential of these genotypes for breeding programs. The traits of greatest importance for dissimilarity were percentage of pulp and of skin, which are easily measured. The clustering structure is related to the collection sites and for breeding programs, genotypes from different sites should be crossed to generate progenies to be tested. Genotypes 'CV5' and 'VT3' should be conserved in genebanks, due to its important agronomic traits.

On the dissimilarity between the algorithms for computing the symmetric energy-momentum tensor in ordinary field theory and general relativity

The recipe used to compute the symmetric energy-momentum tensor in the framework of ordinary field theory bears little resemblance to that used in the context of general relativity, if any. We show that if one stal ts fi om the field equations instead of the Lagrangian density, one obtains a unified algorithm for computing the symmetric energy-momentum tensor in the sense that it can be used for both usual field theory and general relativity.

Unraveling the drivers of community dissimilarity and species extinction in fragmented landscapes

Communities in fragmented landscapes are often assumed to be structured by species extinction due to habitat loss, which has led to extensive use of the species-area relationship (SAR) in fragmentation studies. However, the use of the SAR presupposes that habitat loss leads species to extinction but does not allow for extinction to be offset by colonization of disturbed-habitat specialists. Moreover, the use of SAR assumes that species richness is a good proxy of community changes in fragmented landscapes. Here, we assessed how communities dwelling in fragmented landscapes are influenced by habitat loss at multiple scales; then we estimated the ability of models ruled by SAR and by species turnover in successfully predicting changes in community composition, and asked whether species richness is indeed an informative community metric. To address these issues, we used a data set consisting of 140 bird species sampled in 65 patches, from six landscapes with different proportions of forest cover in the Atlantic Forest of Brazil. We compared empirical patterns against simulations of over 8 million communities structured by different magnitudes of the power-law SAR and with species-specific rules to assign species to sites. Empirical results showed that, while bird community composition was strongly influenced by habitat loss at the patch and landscape scale, species richness remained largely unaffected. Modeling results revealed that the compositional changes observed in the Atlantic Forest bird metacommunity were only matched by models with either unrealistic magnitudes of the SAR or by models ruled by species turnover, akin to what would be observed along natural gradients. We show that, in the presence of such compositional turnover, species richness is poorly correlated with species extinction, and z values of the SAR strongly underestimate the effects of habitat loss. We suggest that the observed compositional changes are driven by each species reaching its individual extinction threshold: either a threshold of forest cover for species that disappear with habitat loss, or of matrix cover for species that benefit from habitat loss.