Comparing Support Vector Machines with Gaussian Kernels to Radial Basis Function Classifiers

The Support Vector (SV) machine is a novel type of learning machine, based on statistical learning theory, which contains polynomial classifiers, neural networks, and radial basis function (RBF) networks as special cases. In the RBF case, the SV algorithm automatically determines centers, weights and threshold such as to minimize an upper bound on the expected test error. The present study is devoted to an experimental comparison of these machines with a classical approach, where the centers are determined by $k$--means clustering and the weights are found using error backpropagation. We consider three machines, namely a classical RBF machine, an SV machine with Gaussian kernel, and a hybrid system with the centers determined by the SV method and the weights trained by error backpropagation. Our results show that on the US postal service database of handwritten digits, the SV machine achieves the highest test accuracy, followed by the hybrid approach. The SV approach is thus not only theoretically well--founded, but also superior in a practical application.

Priors Stabilizers and Basis Functions: From Regularization to Radial, Tensor and Additive Splines

We had previously shown that regularization principles lead to approximation schemes, as Radial Basis Functions, which are equivalent to networks with one layer of hidden units, called Regularization Networks. In this paper we show that regularization networks encompass a much broader range of approximation schemes, including many of the popular general additive models, Breiman's hinge functions and some forms of Projection Pursuit Regression. In the probabilistic interpretation of regularization, the different classes of basis functions correspond to different classes of prior probabilities on the approximating function spaces, and therefore to different types of smoothness assumptions. In the final part of the paper, we also show a relation between activation functions of the Gaussian and sigmoidal type.

Análisis de la influencia de la Cadena Radial “La Luna” en la cultura política de la sociedad ecuatoriana, específicamente en la sociedad quiteña durante el derrocamiento del presidente Lucio Gutiérrez. Periodo del 13 al 20 de abril del 2005

En el presente trabajo se pretendió demostrar la importancia que tiene el ciudadano como agente de veeduría y control de las instancias Estatales, y que a través de los medios se facilita la postulación de las demandas que han sido encubadas y maduradas en espacios de discusión y debate político en instancias locales y regionales. Para esto se escogió el caso de la Rebelión de los Forajidos que tuvo lugar en Quito, Ecuador.

Concordancia interobservador entre ortopedistas de urgencias y cirujanos de codo en fracturas de cúpula radial

El trauma de codo y la fractura de cúpula radial resultan ser un motivo de consulta frecuente en nuestro medio. Es común en nuestro medio que a pesar de tener radiografías de codo en las que se evidencia fractura de cúpula radial se solicite un TAC de codo. Consideramos que el TAC es una herramienta útil en algunos casos de fracturas de la cúpula radial, es decir, no se debe tomar de forma rutinaria el TAC de codo en los pacientes con fracturas de cúpula radial como es usual en nuestro medio. Existen pocos estudios que comparan la concordancia inter observador de las clasificaciones utilizadas para las fracturas de cúpula radial, pero, no existe en la literatura mundial un estudio que evalué las diferencias entre la clasificación de la fractura en base a radiografías y tomografía entre un grupo ortopedistas traumatólogos y sub-especialistas de miembro superior. La realización de este estudio permite elaborar guías de manejo de pacientes con fractura de cúpula radial en la que se racionalizara el uso del TAC de codo.

¿La tomografía de codo cambia la conducta tomada previamente con radiografía en fracturas de cúpula radial? un estudio de concordancia

Las fracturas de la cúpula radial corresponden a un tercio de todas las fracturas del codo en adultos (1), por lo que se toman radiografías o tomografías, sin embargo no existe literatura que soporte la realización de la tomografía como estudio complementario, por lo que preguntamos ¿La tomografía de codo cambia la conducta tomada previamente con radiografía en fracturas de cúpula radial? Se propone un estudio de concordancia, donde se evalúan las radiografías y tomografías por parte de dos cirujanos de codo, de pacientes con diagnóstico de fractura de cúpula radial valorados en urgencias de dos hospitales de tercer nivel de Bogotá desde enero 2011 a enero 2013. Se revisaron 116 historias, 99 cumplieron los criterios, las radiografías fueron revisadas por dos ortopedistas de codo quienes realizaron clasificación de Mason obteniendo un κ 1 (p˂ 0,00), propuesta de tratamiento postradiografía κ 0,934 (p˂ 0,000 IC95% 0.85, 1). Tratamiento post-tomografía de codo κ 0.949 (p˂ 0,00 IC95% 0.867,1). Concordancia intra-ortopedista de la conducta pre y post-tomografía hay cambio en la conducta del ortopedista 1 en 32.6% (κ 0.674 p˂ 0,00 IC95% 0.52, 0.818) y del ortopedista 2 en 36% (κ 0.64 p˂ 0,00 IC95% 0.452, 0.792). Se dividieron los pacientes en Mason I, II, III con concordancia considerable en los pacientes Mason I y III, mientras que en Mason II fue una concordancia aceptable lo que significa que el uso de tomografía de codo cambia la conducta en general pero sobretodo en pacientes clasificados Masson II.

Dehydration does not affect the radial pressures produced by the earthworm Aporrectodea caliginosa

As a soil dries, the earthworms in that soil dehydrate and become less active. Moisture stress may weaken an earthworm, lowering the radial pressure that the animal can produce. This possibility was investigated for the earthworm Aporrectodea caliginosa (Savigny). Pressures were compared for saturated earthworms (worms taken from saturated soil) and stressed earthworms (worms that had been partially dehydrated by leaving them in dry soil). A load cell was used to record the forces that earthworms produced as they moved through artificial burrows (holes that had been drilled through blocks of aluminium or Perspex). The radial pressure was calculated using the forces exerted and the dimensions of the artificial burrows. There was a negative correlation between burrow diameter and radial pressure, although radial pressure was independent of the length of the block through which the earthworms had burrowed. The highest radial pressures were produced by the anterior segments of the animal. Partial dehydration caused the earthworms to become quiescent, but did not decrease the radial pressure that the earthworms produced. It is suggested that coelomic fluid is retained in the anterior segments while the rest of the animal dehydrates. Dehydrated earthworms became lethargic, and we suggest that lethargy is due to the loss of coelomic fluid from the posterior segments. Coelomic fluid is known to be lost through dorsal pores. In burrowing species of earthworm such as Aporrectodea caliginosa, these pores are only present on the posterior segments.

The radial width of a coronal mass ejection between 0.1 and 0.4 AU estimated from the heliospheric imager on STEREO

On 15-17 February 2008, a CME with an approximately circular cross section was tracked through successive images obtained by the Heliospheric Imager (HI) instrument onboard the STEREO-A spacecraft. Reasoning that an idealised flux rope is cylindrical in shape with a circular cross-section, best fit circles are used to determine the radial width of the CME. As part of the process the radial velocity and longitude of propagation are determined by fits to elongation-time maps as 252±5 km/s and 70±5° respectively. With the longitude known, the radial size is calculated from the images, taking projection effects into account. The radial width of the CME, S (AU), obeys a power law with heliocentric distance, R, as the CME travels between 0.1 and 0.4 AU, such that S=0.26 R0.6±0.1. The exponent value obtained is compared to published studies based on statistical surveys of in situ spacecraft observations of ICMEs between 0.3 and 1.0 AU, and general agreement is found. This paper demonstrates the new opportunities provided by HI to track the radial width of CMEs through the previously unobservable zone between the LASCO field of view and Helios in situ measurements.

The accuracy of using the Ulysses result of the spatial invariance of the radial heliospheric field to compute the open solar flux

We survey observations of the radial magnetic field in the heliosphere as a function of position, sunspot number, and sunspot cycle phase. We show that most of the differences between pairs of simultaneous observations, normalized using the square of the heliocentric distance and averaged over solar rotations, are consistent with the kinematic "flux excess" effect whereby the radial component of the frozen-in heliospheric field is increased by longitudinal solar wind speed structure. In particular, the survey shows that, as expected, the flux excess effect at high latitudes is almost completely absent during sunspot minimum but is almost the same as within the streamer belt at sunspot maximum. We study the uncertainty inherent in the use of the Ulysses result that the radial field is independent of heliographic latitude in the computation of the total open solar flux: we show that after the kinematic correction for the excess flux effect has been made it causes errors that are smaller than 4.5%, with a most likely value of 2.5%. The importance of this result for understanding temporal evolution of the open solar flux is reviewed.

Non-radial solar wind flows induced by the motion of interplanetary coronal mass ejections

A survey of the non-radial flows (NRFs) during nearly five years of interplanetary observations revealed the average non-radial speed of the solar wind flows to be �30 km/s, with approximately one-half of the large (>100 km/s) NRFs associated with ICMEs. Conversely, the average non-radial flow speed upstream of all ICMEs is �100 km/s, with just over one-third preceded by large NRFs. These upstream flow deflections are analysed in the context of the large-scale structure of the driving ICME. We chose 5 magnetic clouds with relatively uncomplicated upstream flow deflections. Using variance analysis it was possible to infer the local axis orientation, and to qualitatively estimate the point of interception of the spacecraft with the ICME. For all 5 events the observed upstream flows were in agreement with the point of interception predicted by variance analysis. Thus we conclude that the upstream flow deflections in these events are in accord with the current concept of the large scale structure of an ICME: a curved axial loop connected to the Sun, bounded by a curved (though not necessarily circular)cross section.

Fully complex-valued radial basis function networks for orthogonal least squares regression

We consider a fully complex-valued radial basis function (RBF) network for regression application. The locally regularised orthogonal least squares (LROLS) algorithm with the D-optimality experimental design, originally derived for constructing parsimonious real-valued RBF network models, is extended to the fully complex-valued RBF network. Like its real-valued counterpart, the proposed algorithm aims to achieve maximised model robustness and sparsity by combining two effective and complementary approaches. The LROLS algorithm alone is capable of producing a very parsimonious model with excellent generalisation performance while the D-optimality design criterion further enhances the model efficiency and robustness. By specifying an appropriate weighting for the D-optimality cost in the combined model selecting criterion, the entire model construction procedure becomes automatic. An example of identifying a complex-valued nonlinear channel is used to illustrate the regression application of the proposed fully complex-valued RBF network.

Grey-box radial basis function modelling: the art of incorporating prior knowledge

A basic principle in data modelling is to incorporate available a priori information regarding the underlying data generating mechanism into the modelling process. We adopt this principle and consider grey-box radial basis function (RBF) modelling capable of incorporating prior knowledge. Specifically, we show how to explicitly incorporate the two types of prior knowledge: the underlying data generating mechanism exhibits known symmetric property and the underlying process obeys a set of given boundary value constraints. The class of orthogonal least squares regression algorithms can readily be applied to construct parsimonious grey-box RBF models with enhanced generalisation capability.