Supervised Gaussian Process Latent Variable Model for Dimensionality Reduction

The Gaussian process latent variable model (GP-LVM) has been identified to be an effective probabilistic approach for dimensionality reduction because it can obtain a low-dimensional manifold of a data set in an unsupervised fashion. Consequently, the GP-LVM is insufficient for supervised learning tasks (e. g., classification and regression) because it ignores the class label information for dimensionality reduction. In this paper, a supervised GP-LVM is developed for supervised learning tasks, and the maximum a posteriori algorithm is introduced to estimate positions of all samples in the latent variable space. We present experimental evidences suggesting that the supervised GP-LVM is able to use the class label information effectively, and thus, it outperforms the GP-LVM and the discriminative extension of the GP-LVM consistently. The comparison with some supervised classification methods, such as Gaussian process classification and support vector machines, is also given to illustrate the advantage of the proposed method.

Sensitive determination of Cd and Pb by differential pulse stripping voltammetry with in situ bismuth-modified zeolite doped carbon paste electrodes

Here we investigated the analytical performances of the bismuth-modified zeolite doped carbon paste electrode (BiF-ZDCPE) for trace Cd and Pb analysis. The characteristics of bismuth-modified electrodes were improved greatly via addition of synthetic zeolite into carbon paste. To obtain high reproducibility and sensitivity, optimum experimental conditions for bismuth deposition Were Studied.

Super-hydrophobicity of silica nanoparticles modified with vinyl groups

Super-hydrophobic films with vinyl-modified silica nanoparticles (V-SiOx-NPs) were successfully prepared. The rough surface, which was composed of microstructures of disordered V-SiOx-NPs and nanostructures on the surface of V-SiOx-NPs, rather than the chemical composition devoted to the super-hydrophobicity of film. The relationship between contact angle and diameter of V-SiOx-NPs was then investigated. The sessile contact angles (CA) of films with 150-1600nm V-SiOx-NPs were around 166 regardless the diameter, while the film with 85 nm V-SiOx-NPs had the lowest CA of about 158. The packing manner of V-SiOx-NPs determined the air fraction on the surface and then the CA.

Macroporous fluoropolymeric films templated by silica colloidal assembly: A possible route to super-hydrophobic surfaces

A super-hydrophobic surface was obtained on a three-dimensional (313) polyvinylidene fluoride (PVDF) macroporous film. The porous films were fabricated through self-assembled silica colloidal templates. The apparent water contact angle of the surface can be tuned from 106 degrees to 153 degrees through altering the sintering temperature and the diameter of the colloidal templates. A composite structure of micro-cavities and nanoholes on the PVDF surface was responsible for the super-hydrophobicity. The wettability of the porous surfaces was described by the use of the Cassie-Baxter model and Wenzel's equation.

Fractal roughness characterization of super-ground Mn-Zn ferrite single crystals

The surface of superground Mn-Zn ferrite single crystal may be identified as a self-affine fractal in the stochastic sense. The rms roughness increased as a power of the scale from 10(2) nm to 10(6) nm with the roughness exponent alpha = 0.17 +/- 0.04, and 0.11 +/- 0.06, for grinding feed rate of 15 and 10 mu m/rev, respectively. The scaling behavior coincided with the theory prediction well used for growing self-affine surfaces in the interested region for magnetic heads performance. The rms roughnesses increased with increase in the feed rate, implying that the feed rate is a crucial grinding parameter affecting the supersmooth surface roughness in the machining process.

Estimation of suspended sediment concentrations using Pulse-coherent Acoustic Doppler Profiler (PCADP)

The objective of the study is to investigate the suitability of using Pulse-coherent Acoustic Doppler Profiler (PCADP) to estimate suspended sediment concentration (SSC). The acoustic backscatter intensity was corrected for spreading and absorption loss, then calibrated with OBS and finally converted to SSC. The results show that there is a good correlation between SSC and backscatter intensity with R value of 0.74. The mean relative error is 22.4%. Then the time span of little particle size variation was also analyzed to exclude the influence of size variation. The correlation coefficient increased to 0.81 and the error decreased to 18.9%. Our results suggest that the PCADP can meet the requirement of other professional instruments to estimate SSC with the errors between 20% and 50%, and can satisfy the need of dynamics study of suspended particles.

Deconvolution of overlapped peaks based on the exponentially modified Gaussian model in comprehensive two-dimensional gas chromatography

Comprehensive two-dimensional gas chromatography (GC x GC) has attracted much attention for the analys is of complex samples. Even with a large peak capacity in GC x GC, peak overlapping is often met. In this paper, a new method was developed to resolve overlapped peaks based on the mass conservation and the exponentially modified Gaussian (EMG) model. Linear relationships between the calculated sigma, tau of primary peaks with the corresponding retention time (t(R)) were obtained, and the correlation coefficients were over 0.99. Based on such relationships, the elution profile of each compound in overlapped peaks could be simulated, even for the peak never separated on the second-dimension. The proposed method has proven to offer more accurate peak area than the general data processing method. (c) 2005 Elsevier B.V. All rights reserved.