A hierarchical multiclass support vector machine incorporated with holistic triple learning units

This paper proposes a new hierarchical learning structure, namely the holistic triple learning (HTL), for extending the binary support vector machine (SVM) to multi-classification problems. For an N-class problem, a HTL constructs a decision tree up to a depth of A leaf node of the decision tree is allowed to be placed with a holistic triple learning unit whose generalisation abilities are assessed and approved. Meanwhile, the remaining nodes in the decision tree each accommodate a standard binary SVM classifier. The holistic triple classifier is a regression model trained on three classes, whose training algorithm is originated from a recently proposed implementation technique, namely the least-squares support vector machine (LS-SVM). A major novelty with the holistic triple classifier is the reduced number of support vectors in the solution. For the resultant HTL-SVM, an upper bound of the generalisation error can be obtained. The time complexity of training the HTL-SVM is analysed, and is shown to be comparable to that of training the one-versus-one (1-vs.-1) SVM, particularly on small-scale datasets. Empirical studies show that the proposed HTL-SVM achieves competitive classification accuracy with a reduced number of support vectors compared to the popular 1-vs-1 alternative.

A national survey of the mangement of delirium in UK intensive care units

Background: Delirium is an acute organ dysfunction common amongst patients treated in intensive care units. The associated morbidity and mortality are known to be substantial. Previous surveys have described which screening tools are used to diagnose delirium and which medications are used to treat delirium, but these data are not available for the United Kingdom. Aim: This survey aimed to describe the UK management of delirium by consultant intensivists. Additionally, knowledge and attitudes towards management of delirium were sought. The results will inform future research in this area. Methods: A national postal survey of members of the UK Intensive Care Society was performed. A concise two page questionnaire survey was sent, with a second round of surveys sent to non-respondents after 6 weeks. The questionnaire was in tick-box format. Results: Six hundred and eighty-one replies were received from 1308 questionnaires sent, giving a response rate of 52%. Twenty-five percent of respondents routinely screen for delirium, but of these only 55% use a screening tool validated for use in intensive care. The majority (80%) of those using a validated instrument used the Confusion Assessment Method for the Intensive Care Unit. Hyperactive delirium is treated pharmacologically by 95%; hypoactive delirium is treated pharmacologically by 25%, with haloperidol the most common agent used in both. Over 80% of respondents agreed that delirium prolongs mechanical ventilation and hospital stay and requires active treatment. Conclusions: This UK survey demonstrates screening for delirium is sporadic. Pharmacological treatment is usually with haloperidol in spite of the limited evidence to support this practice. Hypoactive delirium is infrequently treated pharmacologically.

Atomic oxygen formation in a radio-frequency driven micro-atmospheric pressure plasma jet

Atomic oxygen formation in a radio-frequency driven micro-atmospheric pressure plasma jet is investigated using both advanced optical diagnostics and numerical simulations of the dynamic plasma chemistry. Laser spectroscopic measurements of absolute densities of ground state atomic oxygen reveal steep gradients at the interface between the plasma core and the effluent region. Spatial profiles resolving the interelectrode gap within the core plasma indicate that volume processes dominate over surface reactions. Details of the production and destruction processes are investigated in numerical simulations benchmarked by phase-resolved optical emission spectroscopy. The main production mechanisms are electron induced and hence most efficient in the vicinity of the plasma boundary sheath, where electrons are energized. The destruction is driven through chemical heavy particle reactions. The resulting spatial profile of atomic oxygen is relatively flat. The power dependence of the atomic oxygen density obtained by the numerical simulation is in very good agreement with the laser spectroscopic measurements.

Spitzer measurements of atomic and molecular abundances in the Type IIPSN 2005af

We present results based on mid-infrared (3.6-30 mm) observations with the Spitzer Space Telescope of the nearby Type IIP supernova 2005af. We report the first ever detection of the SiO molecule in a Type IIP supernova. Together with the detection of the CO fundamental, this is an exciting finding as it may signal the onset of dust condensation in the ejecta. From a wealth of fine-structure lines we provide abundance estimates for stable Ni, Ar, and Ne that, via spectral synthesis, may be used to constrain nucleosynthesis models.

Atomic ordering in nano-layered FePt: Multiscale Monte Carlo simulation

Nano- and meso-scale simulation of chemical ordering kinetics in nano-layered L1(0)-AB binary intermetallics was performed. In the nano- (atomistic) scale Monte Carlo (MC) technique with vacancy mechanism of atomic migration implemented with diverse models for the system energetics was used. The meso-scale microstructure evolution was, in turn, simulated by means of a MC procedure applied to a system built of meso-scale voxels ordered in particular L1(0) variants. The voxels were free to change the L1(0) variant and interacted with antiphase-boundary energies evaluated within the nano-scale simulations. The study addressed FePt thin layers considered as a material for ultra-high-density magnetic storage media and revealed metastability of the L1(0) c-variant superstructure with monoatomic planes parallel to the (001)-oriented layer surface and off-plane easy magnetization. The layers, originally perfectly ordered in the c-variant, showed discontinuous precipitation of a- and b-L1(0)-variant domains running in parallel with homogeneous disordering (i.e. generation of antisite defects). The domains nucleated heterogeneously on the free monoatomic Fe surface of the layer, grew inwards its volume and relaxed towards an equilibrium microstructure of the system. Two

Atomic ordering in nano-layered FePt

Monte Carlo simulation of chemical ordering kinetics in nano-layered L1₀ AB binary intermetallics was performed. The study addressed FePt thin layers considered as a material for ultra-high-density magnetic storage media and revealed metastability of the L1₀ c-variant superstructure with monoatomic planes parallel to the surface and off-plane easy magnetization. The layers, originally perfectly ordered in a c-variant of the L1₀ superstructure, showed homogeneous disordering running in parallel with a spontaneous re-orientation of the monoatomic planes leading to a mosaic microstructure composed of a- and b-L1₀-variant domains. The domains nucleated heterogeneously on the surface of the layer and grew discontinuously inwards its volume. Finally, the domains relaxed towards an equilibrium microstructure of the system. Two “atomistic-scale” processes: (i) homogeneous disordering and (ii) nucleation of the a- and b-L1₀-variant domains showed characteristic time scales. The same was observed for the domain microstructure relaxation. The discontinuous domain growth showed no definite driving force and proceeded due to thermal fluctuations. The above complex structural evolution has recently been observed experimentally in epitaxially deposited thin films of FePt.

Melt-compounded nanocomposites of titanium dioxide atomic-layer-deposition-coated polyamide and polystyrene powders

Polyamide and polystyrene particles were coated with titanium dioxide films by atomic layer deposition (ALD) and then melt-compounded to form polymer nanocomposites. The rheological properties of the ALD-created nanocomposite materials were characterized with a melt flow indexer, a melt flow spiral mould, and a rotational rheometer. The results suggest that the melt flow properties of polyamide nanocomposites were markedly better than those of pure polyamide and polystyrene nanocomposites. Such behavior was shown to originate in an uncontrollable decrease in the polyamide molecular weight, likely affected by a high thin-film impurity content, as shown in gel permeation chromatography (GPC) and scanning electron microscope (SEM) equipped with an energy-dispersive spectrometer. Transmission electron microscope image showed that a thin film grew on both studied polymer particles, and that subsequent melt-compounding was successful, producing well dispersed ribbon-like titanium dioxide with the titanium dioxide filler content ranging from 0.06 to 1.12wt%. Even though we used nanofillers with a high aspect ratio, they had only a minor effect on the tensile and flexural properties of the polystyrene nanocomposites. The mechanical behavior of polyamide nanocomposites was more complex because of the molecular weight degradation. Our approach here to form polymeric nanocomposites is one way to tailor ceramic nanofillers and form homogenous polymer nanocomposites with minimal work-related risks in handling powder form nanofillers. However, further research is needed to gauge the commercial potential of ALD-created nanocomposite materials. Copyright (C) 2011 John Wiley & Sons, Ltd.

Unlocking synergies between business units: internal value creation at Royal Vopak.

This paper examines the process of creating and exploiting synergies between business units of a multi-unit corporation and the creation of internal value by combining and exploiting knowledge. It offers a framework to create and manage such synergies and undertakes an empirical test through in-depth study across three business units of Royal Vopak, a Dutch-based global multi-unit corporation. Finally, it offers lessons for corporate managers trying to create and manage cross-unit synergies.