LoVis: Local Pattern Visualization for Model Refinement

Linear models are commonly used to identify trends in data. While it is an easy task to build linear models using pre‐selected variables, it is challenging to select the best variables from a large number of alternatives. Most metrics for selecting variables are global in nature, and thus not useful for identifying local patterns. In this work, we present an integrated framework with visual representations that allows the user to incrementally build and verify models in three model spaces that support local pattern discovery and summarization: model complementarity, model diversity, and model representivity. Visual representations are designed and implemented for each of the model spaces. Our visualizations enable the discovery of complementary variables, i.e., those that perform well in modeling different subsets of data points. They also support the isolation of local models based on a diversity measure. Furthermore, the system integrates a hierarchical representation to identify the outlier local trends and the local trends that share similar directions in the model space. A case study on financial risk analysis is discussed, followed by a user study.     

Efficacy of iron fortification compared to iron supplementation among Vietnamese schoolchildren

Background  

Parents of six children are facing a trial on charges of aggravated manslaughter in the care a 5 1/2 month old infant who died suddenly and neglect of their four older children for causing them to be malnourished by feeding them all an exclusively raw foods vegan diet. Both parents declined plea bargains and plan to defend themselves in court.     

Applying Wide & Deep Learning Model for Android Malware Classification

Android malware has exploded in popularity in recent years, due to the platform’s dominance of the mobile market. With the advancement of deep learning technology, numerous deep learning-based works have been proposed for the classification of Android malware. Deep learning technology is designed to handle a large amount of raw and continuous data, such as image content data. However, it is incompatible with discrete features, i.e., features gathered from multiple sources. Furthermore, if the feature set is already well-extracted and sparsely distributed, this technology is less effective than traditional machine learning. On the other hand, a wide learning model can expand the feature set to enhance the classification accuracy. To maximize the benefits of both methods, this study proposes combining the components of deep learning based on multi-branch CNNs (Convolutional Network Neural) with wide learning method. The feature set is evaluated and dynamically partitioned according to its meaning and generalizability to subsets when used as input to the model’s wide or deep component. The proposed model, partition, and feature set quality are all evaluated using the K-fold cross validation method on a composite dataset with three types of features: API, permission, and raw image. The accuracy with Wide and Deep CNN (WDCNN) model is 98.64%, improved by 1.38% compared to RNN (Recurrent Neural Network) model.     

Vapor- and liquid-phase adsorption of alcohol and water in silicalite-1 synthesized in fluoride media

In this work, batch-adsorption experiments and molecular simulations are employed to probe the adsorption of binary mixtures containing ethanol or a linear alkane-1,n-diol solvated in water or ethanol onto silicalite-1. Since the batch-adsorption experiments require an additional relationship to determine the amount of solute (and solvent) adsorbed, as only the bulk liquid reservoir can be probed directly, molecular simulations are used to provide a relationship between solute and solvent adsorption for input to the experimental bulk measurements. The combination of bulk experimental measurements and simulated solute–solvent relationship yields solvent and solute loadings that are self-consistent with simulation alone, and allow for an assessment of the various assumptions made in the literature. At low solution concentrations, the solute loading calculated is independent of the assumption made. At high concentrations, a negligent choice of assumption can lead to systematic overestimation or underestimation of calculated solute loading.     

Interval functional observers for time-delay systems with additive disturbances

This article considers the design of interval functional observers to estimate a linear function of the state vector of time-delay systems subject to both input and output additive disturbances. Two novel functional observers are proposed and designed such that they bound the set of all admissible values of a linear function of the state vector at each instant of time. By contrast to interval observers currently available in the literature, both observers proposed in this article utilize multiple delayed output measurement and have a more general structure. This trade-off feature overcomes some drawbacks in previous work and enables interval functional observers to be designed for a wider class of time-delay systems. Conditions for the existence of interval functional observers are derived and an effective design algorithm for computing unknown observer matrices is provided. Two illustrative examples are given to show the advantages and effectiveness of our design method.     

Two channel digital watermarking for music based on exponential time-spread echo kernel

To robustly and imperceptibly embed copyright information as a watermark into music clips, a two channel digital watermarking system based on the least significant bit (LSB) method and echo hiding (EH) with a new exponential time-spread echo kernel (ETEK) is proposed. By spreading the echo so that it seems like natural room reverberation, as well as the use of two channels to separately watermark by the LSB method and the EH method which have complementary characteristics, the system provides satisfactory sound quality and exhibits both efficient retrieval capacity and robustness against attacks. Experiments show that the proposed system retrieves watermarks with 100% accuracy for all clips and is almost imperceptible. With a set of 42 attacks supplied by the StirMark Benchmark for Audio (SMBA), the proposed system has an average robustness of 0.98 where the ideal value is 1.00, and higher than any conventional methods. With these results, the proposed system is shown to be possible for the application to music distribution over the Internet.     

Magnetoelectric sensor for microtesla magnetic-fields based on (Fe80Co20)78Si12B10/PZT laminates

The magnetoelectric sensor based on (Fe₈₀Co₂₀)₇₈Si₁₂B₁₀/PZT laminates is designed, fabricated and characterized for determining dc and ac magnetic-field strengths as well as field orientations. At low dc magnetic-fields, a ME-voltage response (dV_ME/dH) as high as 2 mV/Oe is achieved. The linear relation V_ME(h_ac) with a slope of dV_ME/dh_ac of 17 mV/Oe shows a great ability to self-powered detecting low ac magnetic-fields. The field orientation can be detected by using the sinusoidal dependence of the magnetoelectric voltage. The sensor is promising not only for microtesla magnetic-field sensing but also for magnetic biosensor applications.     

Reconstructing surfaces by volumetric regularization using radial basis functions

We present a new method of surface reconstruction that generates smooth and seamless models from sparse, noisy, nonuniform, and low resolution range data. Data acquisition techniques from computer vision, such as stereo range images and space carving, produce 3D point sets that are imprecise and nonuniform when compared to laser or optical range scanners. Traditional reconstruction algorithms designed for dense and precise data do not produce smooth reconstructions when applied to vision-based data sets. Our method constructs a 3D implicit surface, formulated as a sum of weighted radial basis functions. We achieve three primary advantages over existing algorithms: (1) the implicit functions we construct estimate the surface well in regions where there is little data, (2) the reconstructed surface is insensitive to noise in data acquisition because we can allow the surface to approximate, rather than exactly interpolate, the data, and (3) the reconstructed surface is locally detailed, yet globally smooth, because we use radial basis functions that achieve multiple orders of smoothness.     

Cell surface properties affect colonisation of raw milk by lactic acid bacteria at the microstructure level

Cell localisation in the matrix microstructure of fermented food is important for microbial activity and ripening. To evaluate the importance of physicochemical factors in these phenomena, the physicochemical behaviour of four strains of lactic acid bacteria were investigated after inoculation of raw milk or milk fractions. To discriminate between physicochemical and metabolic properties, media were (i) rapidly acidified to avoid cell division or (ii) bacteria were allowed to ferment milk. To evaluate the impact of milk components, cells were dispersed in protein or fatty globule fractions. From two Lactococcus lactis subsp. lactis biovar. diacetylactis strains selected from their different surface properties investigated in previous studies, after rapid acidification one was encountered in fatty globules and protein gels whereas the other was found in serum and on the surface of globules. After fermentation, the first one formed colonies in the lipid phase or next to it in the whey and the second formed colonies in the protein gel. Strains forming long chains (Lactobacillus bulgaricus and Streptococcus thermophilus) were always found in contact with fatty globules, where they developed. Our results suggest that localisation and subsequent spatial colonisation depend on the physicochemical properties of cell surfaces, on the structure in chains as well as on the metabolic ability to develop on proteins or lipids.     

新型角分辨XPS电子能谱仪

本文介绍了新型角分辨XPS电子能谱的构造、特点,及其应用。