Scaling the Indian Buffet Process via Submodular Maximization

Inference for latent feature models is inherently difficult as the inference space grows exponentially with the size of the input data and number of latent features. In this work, we use Kurihara & Welling (2008)'s maximization-expectation framework to perform approximate MAP inference for linear-Gaussian latent feature models with an Indian Buffet Process (IBP) prior. This formulation yields a submodular function of the features that corresponds to a lower bound on the model evidence. By adding a constant to this function, we obtain a nonnegative submodular function that can be maximized via a greedy algorithm that obtains at least a one-third approximation to the optimal solution. Our inference method scales linearly with the size of the input data, and we show the efficacy of our method on the largest datasets currently analyzed using an IBP model.

Photoconductive hybrid films via directional self-assembly of C 60 on aligned carbon nanotubes

Hybrid nanostructured materials can exhibit different properties than their constituent components, and can enable decoupled engineering of energy conversion and transport functions. Novel means of building hybrid assemblies of crystalline C 60 and carbon nanotubes (CNTs) are presented, wherein aligned CNT films direct the crystallization and orientation of C 60 rods from solution. In these hybrid films, the C 60 rods are oriented parallel to the direction of the CNTs throughout the thickness of the film. High-resolution imaging shows that the crystals incorporate CNTs during growth, yet grazing-incidence X-ray diffraction (GIXD) shows that the crystal structure of the C 60 rods is not perturbed by the CNTs. Growth kinetics of the C 60 rods are enhanced 8-fold on CNTs compared to bare Si, emphasizing the importance of the aligned, porous morphology of the CNT films as well as the selective surface interactions between C 60 and CNTs. Finally, it is shown how hybrid C 60-CNT films can be integrated electrically and employed as UV detectors with a high photoconductive gain and a responsivity of 10 5 A W -1 at low biases (± 0.5 V). The finding that CNTs can induce rapid, directional crystallization of molecules from solution may have broader implications to the science and applications of crystal growth, such as for inorganic nanocrystals, proteins, and synthetic polymers. © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Measuring EGFR Separations on Cells with ∼10 nm Resolution via Fluorophore Localization Imaging with Photobleaching

Detecting receptor dimerisation and other forms of clustering on the cell surface depends on methods capable of determining protein-protein separations with high resolution in the ∼10-50 nm range. However, this distance range poses a significant challenge because it is too large for fluorescence resonance energy transfer and contains distances too small for all other techniques capable of high-resolution in cells. Here we have adapted the technique of fluorophore localisation imaging with photobleaching to measure inter-receptor separations in the cellular environment. Using the epidermal growth factor receptor, a key cancer target molecule, we demonstrate ∼10 nm resolution while continuously covering the range of ∼10-80 nm. By labelling the receptor on cells expressing low receptor numbers with a fluorescent antagonist we have found inter-receptor separations all the way up from 8 nm to 59 nm. Our data are consistent with epidermal growth factor receptors being able to form homo-polymers of at least 10 receptors in the absence of activating ligands. © 2013 Needham et al.

Stability margins of nonlinear receding-horizon control via inverse optimality

Using the nonlinear analog of the Fake Riccati equation developed for linear systems, we derive an inverse optimality result for several receding-horizon control schemes. This inverse optimality result unifies stability proofs and shows that receding-horizon control possesses the stability margins of optimal control laws. © 1997 Elsevier Science B.V.

Taper-free and kinked germanium nanowires grown on silicon via purging and the two-temperature process

We investigate the growth procedures for achieving taper-free and kinked germanium nanowires epitaxially grown on silicon substrates by chemical vapor deposition. Singly and multiply kinked germanium nanowires consisting of 111 segments were formed by employing a reactant gas purging process. Unlike non-epitaxial kinked nanowires, a two-temperature process is necessary to maintain the taper-free nature of segments in our kinked germanium nanowires on silicon. As an application, nanobridges formed between (111) side walls of V-grooved (100) silicon substrates have been demonstrated. © 2012 IOP Publishing Ltd.

Characterisation of nanostructures via terahertz spectroscopy

We have used terahertz spectroscopy to measure the conductivity and time-resolved photoconductivity of a range of semiconducting nanostructures. This article focuses on our recent terahertz conductivity studies on semiconductor nanowires and single walled carbon nanotubes. © 2010 IEEE.

Enhanced infra-red emission from sub-millimeter microelectromechanical systems micro hotplates via inkjet deposited carbon nanoparticles and fullerenes

In this paper, we demonstrate a micro-inkjet printing technique as a reproducible post-process for the deposition of carbon nanoparticles and fullerene adlayers onto fully CMOS compatible micro-electro-mechanical silicon-on-insulator infrared (IR) light sources to enhance their infrared emission. We show experimentally a significant increase in the infrared emission efficiency of the coated emitters. We numerically validate these findings with models suggesting a dominant performance increase for wavelengths <5.5 μm. Here, the bimodal size distribution in the diameter of the carbon nanoparticles, relative to the fullerenes, is an effective mediator towards topologically enhanced emittance of our miniaturised emitters. A 90% improvement in IR emission power density has been shown which we have rationalised with an increase in the mean thickness of the deposited carbon nanoparticle adlayer. © 2013 AIP Publishing LLC.

On-line policy optimisation of Bayesian spoken dialogue systems via human interaction

A partially observable Markov decision process has been proposed as a dialogue model that enables robustness to speech recognition errors and automatic policy optimisation using reinforcement learning (RL). However, conventional RL algorithms require a very large number of dialogues, necessitating a user simulator. Recently, Gaussian processes have been shown to substantially speed up the optimisation, making it possible to learn directly from interaction with human users. However, early studies have been limited to very low dimensional spaces and the learning has exhibited convergence problems. Here we investigate learning from human interaction using the Bayesian Update of Dialogue State system. This dynamic Bayesian network based system has an optimisation space covering more than one hundred features, allowing a wide range of behaviours to be learned. Using an improved policy model and a more robust reward function, we show that stable learning can be achieved that significantly outperforms a simulator trained policy. © 2013 IEEE.

Controlled amine functionality in self-assembled monolayers via the hidden amine route: chemical and electronic tunability.

A synthetic strategy for fabricating a dense amine functionalized self-assembled monolayer (SAM) on hydroxylated surfaces is presented. The assembly steps are monitored by X-ray photoelectron spectroscopy, Fourier transform infrared- attenuated total reflection, atomic force microscopy, variable angle spectroscopic ellipsometry, UV-vis surface spectroscopy, contact angle wettability, and contact potential difference measurements. The method applies alkylbromide-trichlorosilane for the fabrication of the SAM followed by surface transformation of the bromine moiety to amine by a two-step procedure: S(N)2 reaction that introduces the hidden amine, phthalimide, followed by the removal of the protecting group and exposing the free amine. The use of phthalimide moiety in the process enabled monitoring the substitution reaction rate on the surface (by absorption spectroscopy) and showed first-order kinetics. The simplicity of the process, nonharsh reagents, and short reaction time allow the use of such SAMs in molecular nanoelectronics applications, where complete control of the used SAM is needed. The different molecular dipole of each step of the process, which is verified by DFT calculations, supports the use of these SAMs as means to tune the electronic properties of semiconductors and for better synergism between SAMs and standard microelectronics processes and devices.