Cost Complexity-Based Pruning of Ensemble Classifiers

In this paper we study methods that combine multiple classification models learned over separate data sets. Numerous studies posit that such approaches provide the means to efficiently scale learning to large data sets, while also boosting the accuracy of individual classifiers. These gains, however, come at the expense of an increased demand for run-time system resources. The final ensemble meta-classifier may consist of a large collection of base classifiers that require increased memory resources while also slowing down classification throughput. Here, we describe an algorithm for pruning (i.e., discarding a subset of the available base classifiers) the ensemble meta-classifier as a means to reduce its size while preserving its accuracy and we present a technique for measuring the trade-off between predictive performance and available run-time system resources. The algorithm is independent of the method used initially when computing the meta-classifier. It is based on decision tree pruning methods and relies on the mapping of an arbitrary ensemble meta-classifier to a decision tree model. Through an extensive empirical study on meta-classifiers computed over two real data sets, we illustrate our pruning algorithm to be a robust and competitive approach to discarding classification models without degrading the overall predictive performance of the smaller ensemble computed over those that remain after pruning. Received 30 August 2000 / Revised 7 March 2001 / Accepted in revised form 21 May 2001     

Cerebrospinal fluid biomarkers in clinically isolated syndromes and multiple sclerosis

A panel of three cerebrospinal fluid (CSF) markers for clinically isolated syndromes (CIS) and multiple sclerosis (MS), based on SDS-PAGE, 2-D maps, and immunoblot results, is here proposed. No individual marker has any specificity, though, since they appear in a number of other neurological diseases. However the set of three, with the respective modulation sign (up-regulated or maintained at constant level), appears to be unique for MS. These proteins are: tau protein (levels remaining constant and undistinguishable from controls, contrary to up- and downregulation in other neurological disorders); 14-3-3 protein (strong upregulation of distinct isoforms) and cystatin C (changing in accordance to disease stage and progression). As an additional evidence, one can rely in the pattern of isoforms of 14-3-3, as obtained by 2-D maps and Western blot analysis: this pattern further distinguishes the variation of this protein from other neurological syndromes, notably sporadic Creutzfeldt-Jakob disease (sCJD), motor neuron diseases and other dementias. In contrast, a similar qualitative and quantitative upregulation of 14-3-3 is observed in Guillain-Barré syndrome (GBS), a demyelinating condition affecting the peripheral nervous system. To the best of our knowledge, this is the first time in which such a panel of biomarkers is reported in MS.     

Synchronization Architecture for Multimedia Services in Mobile Radio Environments

The management of multimedia teleservices inthird-generation mobile radio environments is achallenging task due to the multicomposite andmultirequirement nature of this kind of traffic.Following the choice of carrying multimedia information onmultiple bearers over the air interface of a wirelesspersonal communication system, and after the selectionof a TDMA-like multiple access protocol based on the bandwidth reservation, the system isrequired to provide for an interstream synchronizationamong components of the multimedia service. This paperpresents an architecture for guaranteeing that temporal relationships among components at thetransmitter side can be maintained, after crossing theair interface, at the receiver side, regardless ofdifferent bearers' delay constraints and delays sufferedin the buffer of the base station. The attentionis focused on a kind of roughsynchronization to be performed on the system MAC leveland refined by higher protocol layers. Static anddynamic priorities concepts are exploited to achieve the targetsynchronization degree among components. Simulationresults, carried out in a case of a video-audioconnection, confirm the effectiveness of the proposedmechanism.     

Generic polar harmonic transforms for invariant image representation

This paper introduces four classes of rotation-invariant orthogonal moments by generalizing four existing moments that use harmonic functions in their radial kernels. Members of these classes share beneficial properties for image representation and pattern recognition like orthogonality and rotation-invariance. The kernel sets of these generic harmonic function-based moments are complete in the Hilbert space of square-integrable continuous complex-valued functions. Due to their resemble definition, the computation of these kernels maintains the simplicity and numerical stability of existing harmonic function-based moments. In addition, each member of one of these classes has distinctive properties that depend on the value of a parameter, making it more suitable for some particular applications. Comparison with existing orthogonal moments defined based on Jacobi polynomials and eigenfunctions has been carried out and experimental results show the effectiveness of these classes of moments in terms of representation capability and discrimination power.     

Comparison of the Pyrolysis Products of Dichlorodimethylsilane in the Chemical Vapor Deposition of Silicon Carbide on Silica in Hydrogen or Argon

A chemical analysis of the pyrolysis gases and solids formed during the deposition of silicon carbide from the decomposition of dichlorodimethylsilane in argon and hydrogen is reported. Depositions were performed at 1 atm pressure, temperatures from 700° to 1100°C, and a mean residence time of approximately 1 min. The chemical analysis shows that, under reactor conditions, the gases formed were mainly methane, hydrogen, silicon tetrachloride, trichloro-silane, and trichloromethylsilane. The presence of hydrogen chloride was not examined. The use of hydrogen, as a carrier gas, decreased the trichloromethylsilane and solid aerosol (smoke) in the reaction products, compared to that present in the argon system, and increased methane, trichlorosilane, and silicon production. Primarily, silicon and silicon carbide were deposited when hydrogen was used as the carrier gas. When argon was used, a complex mixture of silicon carbide and organosilicon compounds was formed. It is hypothesized that, when hydrogen was used as the carrier gas, silicon carbide formed from chlorosilanes and methane, which were products from the decomposition of dichlorodimethylsilane. These products subsequently reacted to form silicon, which then reacted with methane to form silicon carbide. In argon, however, it is hypothesized that silicon carbide can be formed in two ways: (1) from the pyrolysis of solid organosilicon compounds which are products from the pyrolysis of dichlorodimethylsilane in argon and (2) as the reduction of dichlorodimethylsilane to chlorosilanes and methane, caused by the hydrogen produced from the pyrolysis of dichlorodimethylsilane in argon.     

The influence of extent of polymerization on the thermal behavior of an epoxy-based polyether resin

A number of studies have been reported in the literature on the polymerization and thermal decomposition of epoxide resins. Lee¹ and Anderson² have both studied the thermal decomposition of epoxy resins, and they concluded that the characteristic exothermic peak (which can occur anywhere between 300° and 400°C) is caused at least partially by some reaction of the epoxide group. We have been investigating the thermal decomposition of an aromatic polyether resin which is produced by curing the diglycidyl ether of bisphenol A (Epon 825) with the catalytic agent trimethoxyboroxine (Fig. 1). DTA studies of the polyether in an inert atmosphere of N₂ showed exothermic peaks at approximately 390°, 430°, and 470°C, with the major exotherm being the one at 430°C. Our investigation has shown the important role played by low molecular weight epoxides in these exothermic reactions.     

Frying performance of a sunflower/palm oil blend in comparison with pure palm oil

The aim of this study was to test the performance of a vegetable oil blend formulated as alternative to pure palm oil as frying medium. For this purpose, the evolution of many analytical parameters (free acidity, spectrophotometric indices, total polar components, fatty acid composition, short‐chain fatty acids, tocopherol and tocotrienol content and composition, color, flavor evaluated by means of an electronic nose) of the selected blend (sunflower/palm oil 65 : 35 vol/vol) has been monitored during a prolonged frying process (8 h discontinuous frying without oil replenishment) in comparison to pure palm oil. Sensory attributes of the fried food were also evaluated. The blend proved to keep qualitative parameters comparable to those shown by palm oil during the prolonged frying process. Even if some oxidation indices, such as spectrophotometric indices, short‐chain fatty acids and total polar components, increased faster in the blend, it showed a higher tocopherol content and a lower increment in free fatty acids as compared to pure palm oil. Chips fried in the two oils did not show significantly different sensory profiles.     

On Hardiness and Other Pathways to Resilience.

Comments on George Bonanno's article entitles Loss, Trauma, and Human Resilience: Have We Underestimated the Human Capacity to Thrive After Extremely Aversive Events? (see record 2004-10043-003). The author of this comment notes that Bonanno's article is a compelling analysis of several different reactions to extremely aversive events. With regard to major stressors, such as the death of a loved one or the experience a life-threatening circumstance, there is, notably, not only the psychopathological breakdown (e.g., posttraumatic stress disorder [PTSD] or depressive disorder) that has been emphasized but also the less recognized resilient response, in which there is little or no loss of functioning. Whereas the emphasis in the breakdown response is understandably on treatment that facilitates recovery, the resilient response may require little but the person's own ongoing efforts in life to continue effectively. Indeed, Bonanno suggested that the typical treatment efforts following traumatic events might actually undermine the resilient person's adjustment efforts. Bonanno's conceptualization and supportive research evidence advance our understanding of individuals' reactions to massive stressors. Further, Bonanno (2004) argued that there are multiple pathways to resilience under stress, and he identified personality hardiness as one of them. His presentation of hardiness accurately emphasizes its interrelated attitudes of commitment (rather than alienation), control (rather than powerlessness), and challenge (rather than threat) but does not cover how this concept has evolved over the last 25 years. Coordinating theory, research, and practice over this period of time, the author has concluded (Maddi, 2002) that hardy attitudes amount to the courage and motivation to face stressors accurately (rather than to deny or catastrophize them). This courage and motivation lead to coping by problem solving rather than by avoiding and to interacting with others by giving and getting assistance and encouragement rather than by striking out or overprotecting. Thus, personality hardiness is emerging as a pattern of attitudes and actions that helps in transforming stressors from potential disasters into growth opportunities. The author concludes that Bonanno's exciting position has opened the way for such comparative analytic research through which researchers can further their current knowledge concerning how resilience under stress comes about. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

On-line genetic design of anti-windup unstructured controllers for electric drives with variable load

In this paper, we describe an evolutionary design procedure for discrete-time anti-windup controller for electrical drives. Using a genetic algorithm devised to test and compare controllers of different orders, we search for the discrete anti-windup controller achieving the optimal compromise of weighted cost and performance indices. The search is performed on-line, on the physical hardware, by continuously downloading and testing new solutions on a microprocessor running the control algorithms in real time. The controller obtained through genetic search significantly outperforms alternative schemes obtained with conventional design techniques.     

Analysis techniques for detection of IM broken rotor bars after supply disconnection

This paper presents some analysis techniques of the space vector of voltages induced in the stator windings after supply disconnection, to detect broken rotor bars in squirrel-cage induction machines. When the motor is disconnected from the supply no currents flow in the stator windings and the voltages measurable at its terminals are due to flux produced by rotor currents. When the rotor is healthy, the voltages measured at motor terminals are almost sinusoidal because of the symmetry of rotor windings. When there are broken rotor bars, the magnetomotive force generated by rotor windings is distorted, and some particular harmonics, contained in the voltages induced in the stator windings, increase their amplitudes. The diagnostic technique is based on monitoring these voltage harmonics and analyzing the space vector of the voltages induced in the stator windings via MUSIC pseudospectrum and short-time MUSIC (STMUSIC) time-frequency pseudorepresentation. The MUSIC algorithm is based on the eigen analysis of the autocorrelation matrix, and permits us to evidence the principal harmonic frequencies of the signal and decrease the noise influence, thus allowing a better detection of the broken rotor bars. The results obtained using MUSIC and STMUSIC algorithm have been compared experimentally with those obtained by fast Fourier transform (FFT) and short-time FFT, respectively, and two different sized induction motors have been tested, to demonstrate the superiority of the former approach. Differently from most of the diagnostic techniques already proposed in the technical literature, the proposed approach is effective regardless of the load condition of the machine, source characteristics, and iron saturation.