Bayesian Structured Prediction Using Gaussian Processes

We introduce a conceptually novel structured prediction model, GPstruct, which is kernelized, non-parametric and Bayesian, by design. We motivate the model with respect to existing approaches, among others, conditional random fields (CRFs), maximum margin Markov networks (M3N), and structured support vector machines (SVMstruct), which embody only a subset of its properties. We present an inference procedure based on Markov Chain Monte Carlo. The framework can be instantiated for a wide range of structured objects such as linear chains, trees, grids, and other general graphs. As a proof of concept, the model is benchmarked on several natural language processing tasks and a video gesture segmentation task involving a linear chain structure. We show prediction accuracies for GPstruct which are comparable to or exceeding those of CRFs and SVMstruct.

Hierarchically structured nanocarbon electrodes for flexible solid lithium batteries

The ever increasing demand for storage of electrical energy in portable electronic devices and electric vehicles is driving technological improvements in rechargeable batteries. Lithium (Li) batteries have many advantages over other rechargeable battery technologies, including high specific energy and energy density, operation over a wide range of temperatures (-40 to 70. °C) and a low self-discharge rate, which translates into a long shelf-life (~10 years) [1]. However, upon release of the first generation of rechargeable Li batteries, explosions related to the shorting of the circuit through Li dendrites bridging the anode and cathode were observed. As a result, Li metal batteries today are generally relegated to non-rechargeable primary battery applications, because the dendritic growth of Li is associated with the charging and discharging process. However, there still remain significant advantages in realizing rechargeable secondary batteries based on Li metal anodes because they possess superior electrical conductivity, higher specific energy and lower heat generation due to lower internal resistance. One of the most practical solutions is to use a solid polymer electrolyte to act as a physical barrier against dendrite growth. This may enable the use of Li metal once again in rechargeable secondary batteries [2]. Here we report a flexible and solid Li battery using a polymer electrolyte with a hierarchical and highly porous nanocarbon electrode comprising aligned multiwalled carbon nanotubes (CNTs) and carbon nanohorns (CNHs). Electrodes with high specific surface area are realized through the combination of CNHs with CNTs and provide a significant performance enhancement to the solid Li battery performance. © 2013 Elsevier Ltd.

Revealing lithium-silicide phase transformations in nano-structured silicon-based lithium ion batteries via in situ NMR spectroscopy.

Nano-structured silicon anodes are attractive alternatives to graphitic carbons in rechargeable Li-ion batteries, owing to their extremely high capacities. Despite their advantages, numerous issues remain to be addressed, the most basic being to understand the complex kinetics and thermodynamics that control the reactions and structural rearrangements. Elucidating this necessitates real-time in situ metrologies, which are highly challenging, if the whole electrode structure is studied at an atomistic level for multiple cycles under realistic cycling conditions. Here we report that Si nanowires grown on a conducting carbon-fibre support provide a robust model battery system that can be studied by (7)Li in situ NMR spectroscopy. The method allows the (de)alloying reactions of the amorphous silicides to be followed in the 2nd cycle and beyond. In combination with density-functional theory calculations, the results provide insight into the amorphous and amorphous-to-crystalline lithium-silicide transformations, particularly those at low voltages, which are highly relevant to practical cycling strategies.