Bayesian network data imputation with application to survival tree analysis

Retrospective clinical datasets are often characterized by a relatively small sample size and many missing data. In this case, a common way for handling the missingness consists in discarding from the analysis patients with missing covariates, further reducing the sample size. Alternatively, if the mechanism that generated the missing allows, incomplete data can be imputed on the basis of the observed data, avoiding the reduction of the sample size and allowing methods to deal with complete data later on. Moreover, methodologies for data imputation might depend on the particular purpose and might achieve better results by considering specific characteristics of the domain. The problem of missing data treatment is studied in the context of survival tree analysis for the estimation of a prognostic patient stratification. Survival tree methods usually address this problem by using surrogate splits, that is, splitting rules that use other variables yielding similar results to the original ones. Instead, our methodology consists in modeling the dependencies among the clinical variables with a Bayesian network, which is then used to perform data imputation, thus allowing the survival tree to be applied on the completed dataset. The Bayesian network is directly learned from the incomplete data using a structural expectation–maximization (EM) procedure in which the maximization step is performed with an exact anytime method, so that the only source of approximation is due to the EM formulation itself. On both simulated and real data, our proposed methodology usually outperformed several existing methods for data imputation and the imputation so obtained improved the stratification estimated by the survival tree (especially with respect to using surrogate splits).

Assessment of Polycyclic Aromatic Hydrocarbons in an Urban Soil Dataset

We compare a suite of Polycyclic Aromatic Hydrocarbons (Parent PAHs) in soils and air across an urban area (Belfast UK). Isomeric PAH ratios suggest that soil PAHs are mainly from a combustion source. Fugacity modelling across a range of soil temperatures predicts that four ring and larger PAHs from pyrene to indeno[1,2,3–cd]pyrene all partition strongly (>98%) to the soil compartment. This modelling also implies that these PAHs do not experience losses through partitioning to other phases (air, water) due to soil temperature effects. Such modelling may help in understanding the overall contaminantdistribution in soils. The air and soil data together with modelling suggests that care must be taken when considering isomeric ratios of compounds with mass lighter than 178 (i.e. phenanthrene and anthracene) in the soil phase. Comparison of duplicate and replicate samples suggest that field sampling of duplicates dominates uncertainty and validated methodologies for selection of field duplicates and lab splitting are required. As the urban soil four ring PAHs are at equilibrium in the soil phase, and have characteristic ratios that are dominated by a combustion source that is a single controlling factor over spatial distribution, methods that calculate background concentrations can be compared.

High-resolution electron energy-loss spectroscopy of Ba Ti O 3∕ Sr Ti O 3 multilayers

The dielectric properties of BaTiO3 thin films and multilayers are different from bulk materials because of nanoscale dimensions, interfaces, and stress-strain conditions. In this study, BaTiO3/SrTiO3 multilayers deposited on SrTiO3 substrates by pulsed laser deposition have been investigated by high-energy-resolution electron energy-loss spectroscopy. The fine structures in the spectra are discussed in terms of crystal-field splitting and the internal strain. The crystal-field splitting of the BaTiO3 thin layer is found to be a little larger than that of bulk BaTiO3, which has been interpreted by the presence of the internal strain induced by the misfit at the interface. This finding is consistent with the lattice parameters of the BaTiO3 thin layer determined by the selected area diffraction pattern. The near-edge structure of the oxygen K edge in BaTiO3 thin layers and in bulk BaTiO3 are simulated by first-principle self-consistent full multiple-scattering calculations. The results of the simulations are in a good agreement with the experimental results. Moreover, the aggregation of oxygen vacancies at the rough BaTiO3/SrTiO3 interface is indicated by the increased [Ti]/[O] element ratio, which dominates the difference of dielectric properties between BaTiO3 layer and bulk materials.

Stable Isolated Metal Atoms as Active Sites for Photocatalytic Hydrogen Evolution

The process of using solar energy to split water to produce hydrogen assisted by an inorganic semiconductor is crucial for solving our energy crisis and environmental problems in the future. However, most semiconductor photocatalysts would not exhibit excellent photocatalytic activity without loading suitable co-catalysts. Generally, the noble metals have been widely applied as co-catalysts, but always agglomerate during the loading process or photocatalytic reaction. Therefore, the utilization efficiency of the noble co-catalysts is still very low on a per metal atom basis if no obvious size effect exists, because heterogeneous catalytic reactions occur on the surface active atoms. Here, for the first time, we have synthesized isolated metal atoms (Pt, Pd, Rh, or Ru) stably by anchoring on TiO2, a model photocatalystic system, by a facile one-step method. The isolated metal atom based photocatalysts show excellent stability for H-2 evolution and can lead to a 6-13-fold increase in photocatalytic activity over the metal clusters loaded on TiO2 by the traditional method. Furthermore, the configurations of isolated atoms as well as the originality of their unusual stability were analyzed by a collaborative work from both experiments and theoretical calculations.

Unidirectional suppression of hydrogen oxidation on oxidized platinum clusters

Solar-driven water splitting to produce hydrogen may be an ideal solution for global energy and environment issues. Among the various photocatalytic systems, platinum has been widely used to co-catalyse the reduction of protons in water for hydrogen evolution. However, the undesirable hydrogen oxidation reaction can also be readily catalysed by metallic platinum, which limits the solar energy conversion efficiency in artificial photosynthesis. Here we report that the unidirectional suppression of hydrogen oxidation in photocatalytic water splitting can be fulfilled by controlling the valence state of platinum; this platinum-based cocatalyst in a higher oxidation state can act as an efficient hydrogen evolution site while suppressing the undesirable hydrogen back-oxidation. The findings in this work may pave the way for developing other high-efficientcy platinum-based catalysts for photocatalysis, photoelectrochemistry, fuel cells and water-gas shift reactions.

Active sites on hydrogen evolution photocatalyst

Solar hydrogen production assisted with semiconductor materials is a promising way to provide alternative energy sources in the future. Such a photocatalytic reaction normally takes place on the active sites of the catalysts surface, and the identification of the active sites is crucial for understanding the photocatalytic reaction mechanism and further improving the photocatalytic efficiency. However, the active sites of model catalysts are still largely disputed because of their structural complexity. Conventionally, H-2 evolution from solar water splitting over Pt/TiO2 is widely deemed to take place on metallic Pt nanoparticles. Oppositely, we report through a combined experimental and theoretical approach, that metallic Pt nanoparticles have little contribution to the activity of photocatalytic H-2 evolution; the oxidized Pt species embedded on the TiO2 surface are the key active sites and primarily responsible for the activity of the hydrogen evolution Pt/TiO2 photocatalyst.

The application of a novel fluidised photo reactor under UV-Visible and natural solar irradiation in the photocatalytic generation of hydrogen

With advancements in the development of visible light responsive catalysts for H2 production frequently being reported, photocatalytic water splitting has become an attractive method as a potential ‘solar fuel generator’. The development of novel photo reactors which can enhance the potential of such catalyst, however, is rarely reported. This is particularly important as many reactor configurations are mass transport limited, which in term limits the efficiency of more effective photocatalysts in larger scale applications. This paper describes the performance of a novel fluidised photo reactor for the production of H2 over two catalysts under UV-Visible light and natural solar illumination. Catalysts Pt-C3N4 and NaTaO3.La were dispersed in the reactor and the rate of H2 was determined by GC-TCD analysis of the gas headspace. The unit was an annular reactor constructed from stainless steel 316 and quartz glass with a propeller located in the base to control fluidisation of powder catalysts. Reactor properties such as propeller rotational speed were found to enhance the photo activity of the system through the elimination of mass transport limitations and increasing light penetration. The optimum conditions for H2 evolution were found to be a propeller rotational speed of 1035 rpm and 144 W of UV-Visible irradiation, which produced a rate of 89 µmol h-1 g-1 over Pt-C3N4. Solar irradiation was provided by the George Ellery Hale Solar Telescope, located at the California Institute of Technology.

Three-dimensional porous carbon nanofiber networks decorated with cobalt-based nanoparticles: A robust electrocatalyst for efficient water oxidation

Electrochemical water splitting used for generating hydrogen has attracted increasingly attention due to energy and environmental issues. It is a major challenge to design an efficient, robust and inexpensive electrocatalyst to achieve preferable catalytic performance. Herein, a novel three-dimensional (3D) electrocatalyst was prepared by decorating nanostructured biological material-derived carbon nanofibers with in situ generated cobalt-based nanospheres (denoted as CNF@Co) through a facile approach. The interconnected porous 3D networks of the resulting CNF@Co catalyst provide abundant channels and interfaces, which remarkably favor both mass transfer and oxygen evolution. The as-prepared CNF@Co shows excellent electrocatalytic activity towards the oxygen evolution reactions with an onset potential of about 0.445 V vs. Ag/AgCl. It only needs a low overpotential of 314 mV to achieve a current density of 10 mA/cm² in 1.0 M KOH. Furthermore, the CNF@Co catalyst exhibits excellent stability towards water oxidation, even outperforming commercial IrO<inf>2</inf> and RuO<inf>2</inf> catalysts.

Validation of a 3D damage model for predicting the response of composite structures under crushing loads

A 3D intralaminar continuum damage mechanics based material model, combining damage mode interaction and material nonlinearity, was developed to predict the damage response of composite structures undergoing crush loading. This model captures the structural response without the need for calibration of experimentally determined material parameters. When used in the design of energy absorbing composite structures, it can reduce the dependence on physical testing. This paper validates this model against experimental data obtained from the literature and in-house testing. Results show that the model can predict the force response of the crushed composite structures with good accuracy. The simulated energy absorption in each test case was within 12% of the experimental value. Post-crush deformation and the damage morphologies, such as ply splitting, splaying and breakage, were also accurately reproduced. This study establishes the capability of this damage model for predicting the responses of composite structures under crushing loads.