Tracking mobile users in wireless networks via semi-supervised colocalization

Recent years have witnessed the growing popularity of sensor and sensor-network technologies, supporting important practical applications. One of the fundamental issues is how to accurately locate a user with few labeled data in a wireless sensor network, where a major difficulty arises from the need to label large quantities of user location data, which in turn requires knowledge about the locations of signal transmitters or access points. To solve this problem, we have developed a novel machine learning-based approach that combines collaborative filtering with graph-based semi-supervised learning to learn both mobile users' locations and the locations of access points. Our framework exploits both labeled and unlabeled data from mobile devices and access points. In our two-phase solution, we first build a manifold-based model from a batch of labeled and unlabeled data in an offline training phase and then use a weighted k-nearest-neighbor method to localize a mobile client in an online localization phase. We extend the two-phase colocalization to an online and incremental model that can deal with labeled and unlabeled data that come sequentially and adapt to environmental changes. Finally, we embed an action model to the framework such that additional kinds of sensor signals can be utilized to further boost the performance of mobile tracking. Compared to other state-of-the-art systems, our framework has been shown to be more accurate while requiring less calibration effort in our experiments performed on three different testbeds.     

Feature Activated Molecular Dynamics: Parallelization and Application to Systems with Globally Varying Mechanical Fields

A recently developed hybrid molecular dynamics method (Feature Activated Molecular Dynamics, or FAMD), which was originally designed to extend the scope of certain types of molecular dynamics simulations, is extended here in two ways. First, the method is modified to execute on parallel computer architectures using the MPI communication interface. The parallel FAMD algorithm is demonstrated to be computationally efficient and to substantially increase the length scales accessible with molecular dynamics. The performance of the parallel algorithm is demonstrated using a crystalline system containing 1× 10⁶ atoms, in which 1000 supersaturated self-interstitials are introduced and allowed to aggregate for about 4 ns. In the second part of this paper, the FAMD method is applied to problems in which spatio-temporally varying stress fields are present throughout the simulation cell. In particular, we consider the evolution of a spherical void in a hydrostatically stressed silicon crystal and show that the method can capture the extremely rapid void cavitation dynamics following material failure. Once again, the FAMD approach is demonstrated to provide substantial computational advantages over standard molecular dynamics.     

Extended Spectral Regression for efficient scene recognition

This paper proposes a novel method based on Spectral Regression (SR) for efficient scene recognition. First, a new SR approach, called Extended Spectral Regression (ESR), is proposed to perform manifold learning on a huge number of data samples. Then, an efficient Bag-of-Words (BOW) based method is developed which employs ESR to encapsulate local visual features with their semantic, spatial, scale, and orientation information for scene recognition. In many applications, such as image classification and multimedia analysis, there are a huge number of low-level feature samples in a training set. It prohibits direct application of SR to perform manifold learning on such dataset. In ESR, we first group the samples into tiny clusters, and then devise an approach to reduce the size of the similarity matrix for graph learning. In this way, the subspace learning on graph Laplacian for a vast dataset is computationally feasible on a personal computer. In the ESR-based scene recognition, we first propose an enhanced low-level feature representation which combines the scale, orientation, spatial position, and local appearance of a local feature. Then, ESR is applied to embed enhanced low-level image features. The ESR-based feature embedding not only generates a low dimension feature representation but also integrates various aspects of low-level features into the compact representation. The bag-of-words is then generated from the embedded features for image classification. The comparative experiments on open benchmark datasets for scene recognition demonstrate that the proposed method outperforms baseline approaches. It is suitable for real-time applications on mobile platforms, e.g. tablets and smart phones.     

Amphiphilic semiconducting copolymer as compatibility layer for printing polyelectrolyte-gated OFETs

We report a method for inkjet-printing an organic semiconductor layer on top of the electrolyte insulator layer in polyelectrolyte-gated OFETs by using a surface modification treatment to overcome the underlying wettability problem at this interface. The method includes depositing an amphiphilic diblock copolymer (P3HT-b-PDMAEMA). This material is designed to have one set of blocks that mimics the hydrophobic properties of the semiconductor (poly(3-hexylthiophene) or P3HT), while the other set of blocks include polar components that improve adhesion to the polyelectrolyte insulator. Contact angle measurements, atomic force microscopy, and X-ray photoelectron spectroscopy confirm formation of the desired surface modification film. Successful inkjet printing of a smooth semiconductor layer allows us to manufacture complete transistor structures that exhibit low-voltage operation in the range of 1 V.     

Process development of amorphous silicon/crystalline silicon solar cells

We have already investigated some crucial limiting process steps of the amorphous silicon (a-Si)/crystalline silicon (c-Si) solar cell technology and some specific characterization tools of the ultrathin amorphous material used in devices. In this work, we focus our attention particularlyon the technology of the ITO front contact fabrication, that also is used as an antireflective coating. It is pointed out that this layer acts as a barrier layer against the diffusion of metal during the annealing treatments of the front contact grid. The criteria of the selection of the metal to be used to obtain good performance of the grid and the deposition methods best suited to the purpose are shown. We were able to fabricate low temperature heterojunction solar cells based p-type Czochralski silicon, and a conversion efficiency of 14.7% on 3.8 cm² area was obtained without back surface field and texturization.     

Growth of E18 rat hippocampal neuronal cells on nanostructured surfaces for enhanced electrical stimulation and recording

We have prepared gold nanowire arrays inside nanoporous alumina templates with the goal towards neuronal interfacing and electrical recording from neurons. We have investigated biofunctionalization of such gold nanowire arrays (GNWs) and gold nanofilm (GNF) platforms to understand its impact on neuronal attachment and growth. Poly-D-Lysine (PDL) was coated on the nano-templates surfaces for adhesion of neurons which also enhanced the neuronal growth. Optical microscopy and scanning electron microscopy images revealed strong affinity and improved growth of neurons on PDL-coated surfaces. Such results will impact future investigation of stimulation and recording of electrical activity on nanoscale surfaces.     

A geometric approach to multiple-channel signal detection

The paper introduces the generalized coherence (GC) estimate and examines its application as a statistic for detecting the presence of a common but unknown signal on several noisy channels. The GC estimate is developed as a natural generalization of the magnitude-squared coherence (MSC) estimate-a widely used statistic for nonparametric detection of a common signal on two noisy channels. The geometrical nature of the GC estimate is exploited to derive its distribution under the H₀ hypothesis that the data channels contain independent white Gaussian noise sequences. Detection thresholds corresponding to a range of false alarm probabilities are calculated from this distribution. The relationship of the H₀ distribution of the GC estimate to that of the determinant of a complex Wishart-distributed matrix is noted. The detection performance of the three-channel GC estimate is evaluated by simulation using a white Gaussian signal sequence in white Gaussian noise. Its performance is compared with that of the multiple coherence (MC) estimate, another nonparametric multiple-channel detection statistic. The GC approach is found to provide better detection performance than the MC approach in terms of the minimum signal-to-noise ratio on all data channels necessary to achieve desired combinations of detection and false alarm probabilities     

Investigation of point defect clusters in silicon using parallel molecular dynamics

A parallel molecular dynamics algorithm is presented for computingconfigurations of relatively large defects in crystalline silicon, as modelledby the Stillinger–Weber (SW) three-body interatomic potential. Thealgorithm is based on a partitioning of physical space among the N processorswith atoms migrating freely between the partitions. Implementation on aneight-processor IBM SP2 computer shows the increased efficiency withsimulation size expected because of the increased computational load perprocessor relative to communication overhead. The parallel efficiency reached70% for 21 952 atoms. Calculations are presented for the thermodynamics offormation of interstitial and vacancy clusters containing up to seven pointdefects. The clusters were relaxed within a host lattice of about 3000 siliconatoms subjected to periodic boundary conditions. Free energies of formationfor temperatures 500 K T 1600 K were computed using thermodynamicintegration. Computed equilibrium distributions for these clusters show ashift to the larger species at lower temperatures, as expected. The SWpotential predicts greater driving forces for interstitial aggregation thanvacancy aggregation across the entire temperature range. Model calculationsfor a large vacancy cluster are also presented to demonstrate the utility ofthe algorithm for exploring very large defects in silicon.     

Communication-efficient distributed multi-task learning with matrix sparsity regularization

Machine Learning - This work focuses on distributed optimization for multi-task learning with matrix sparsity regularization. We propose a fast communication-efficient distributed optimization...     

The mechanical and thermal properties of graphitic carbon nitride (g-C3N4)-based epoxy composites

Numerous ways to reinforce epoxy resin and improve its thermomechanical properties have been attempted using organic and inorganic nanoparticles. In this paper, graphitic carbon nitride (g-C₃N₄) nanoparticles were synthesized and used to improve the mechanical properties and thermal stability of epoxy composites. The g-C₃N₄ was synthesized from cheap melamine powder using a simple one-step thermal treatment, then was used to reinforce the resin at different weight percentages (wt%). X-ray diffraction, scanning electron microscopy (SEM), and Fourier infrared spectroscopy were used to characterize the g-C₃N₄ and ensure its successful synthesis by studying the changes in its crystal structure, morphology, and chemical structure. The filler was dispersed in the resin using a combination of ultrasonication and high shear mixing. The results showed that the mechanical properties were optimum when 0.5 wt% g-C₃N₄ was used. The tensile strength and fracture toughness of the resulting epoxy composite improved by 21.8% and 77.3%, respectively. SEM was used to investigate the morphologies of cracks formed in epoxy composite specimens after the tensile testing. The SEM micrographs of the fracture surface showed a transition from a brittle to a rough morphology, signifying the enhancement in the composites' toughness. Thermogravimetric analysis showed a good improvement in degradation temperature of up to 8.86% while dynamic mechanical analysis showed that the incorporation of g-C₃N₄ did not affect the material's glass transition temperature.