Abstract: This work deals with aspects of support vector machine learning for large-scale data mining tasks. Based on a decomposition algorithm for support vector machine training that can be run in serial as well as shared memory parallel mode we introduce a transformation of the training data that allows for the usage of an expensive generalized kernel without additional costs. We present experiments for the Gaussian kernel, but usage of other kernel functions is possible, too. In order to further speed up the decomposition algorithm we analyze the critical problem of working set selection for large training data sets. In addition, we analyze the influence of the working set sizes onto the scalability of the parallel decomposition scheme. Our tests and conclusions led to several modifications of the algorithm and the improvement of overall support vector machine learning performance. Our method allows for using extensive parameter search methods to optimize classification accuracy.
Abstract: The Bulgarian natural expanded mineral obtained from Bentonite AD perlite (A deposit of "The Broken Mountain" for perlite mining, near by the village of Vodenicharsko, in the municipality of Djebel), was loaded with silver (as ion form - Ag+ 2 and 5 wt% by the incipient wetness impregnation method), and as atomic silver - Ag0 using Tollen-s reagent (silver mirror reaction). Some physicochemical characterization of the samples are provided via: DC arc-AES, XRD, DR-IR and UV-VIS. The aim of this work was to obtain and test the silver-loaded catalyst for ozone decomposition. So the samples loaded with atomic silver show ca. 80% conversion of ozone 20 minutes after the reaction start. Then conversion decreases to ca. 20 % but stay stable during the prolongation of time.
Abstract: Neural processors have shown good results for
detecting a certain character in a given input matrix. In this paper, a
new idead to speed up the operation of neural processors for character
detection is presented. Such processors are designed based on cross
correlation in the frequency domain between the input matrix and the
weights of neural networks. This approach is developed to reduce the
computation steps required by these faster neural networks for the
searching process. The principle of divide and conquer strategy is
applied through image decomposition. Each image is divided into
small in size sub-images and then each one is tested separately by
using a single faster neural processor. Furthermore, faster character
detection is obtained by using parallel processing techniques to test the
resulting sub-images at the same time using the same number of faster
neural networks. In contrast to using only faster neural processors, the
speed up ratio is increased with the size of the input image when using
faster neural processors and image decomposition. Moreover, the
problem of local subimage normalization in the frequency domain is
solved. The effect of image normalization on the speed up ratio of
character detection is discussed. Simulation results show that local
subimage normalization through weight normalization is faster than
subimage normalization in the spatial domain. The overall speed up
ratio of the detection process is increased as the normalization of
weights is done off line.
Abstract: Our work is part of the heterogeneous data
integration, with the definition of a structural and semantic mediation
model. Our aim is to propose architecture for the heterogeneous
sources metadata mediation, represented by XML, RDF and RuleML
models, providing to the user the metadata transparency. This, by
including data structures, of natures fundamentally different, and
allowing the decomposition of a query involving multiple sources, to
queries specific to these sources, then recompose the result.
Abstract: Global temperature had increased by about 0.5oC over
the past century, increasing temperature leads to a loss or a decrease
of soil organic matter (SOM). Whereas soil organic matter in many
tropical soils is less stable than that of temperate soils, and it will be
easily affected by climate change. Therefore, conservation of soil
organic matter is urgent issue nowadays. This paper presents the
effect of different doses (5%, 15%) of Ca-type zeolite in conjunction
with organic manure, applied to soil samples from Philippines,
Paraguay and Japan, on the decomposition resistance of soil organic
matter under high temperature. Results showed that a remain or
slightly increase the C/N ratio of soil. There are an increase in
percent of humic acid (PQ) that extracted with Na4P2O7. A decrease
of percent of free humus (fH) after incubation was determined. A
larger the relative color intensity (RF) value and a lower the color
coefficient (6logK) value following increasing zeolite rates leading
to a higher degrees of humification. The increase in the aromatic
condensation of humic acid (HA) after incubation, as indicates by the
decrease of H/C and O/C ratios of HA. This finding indicates that the
use of zeolite could be beneficial with respect to SOM conservation
under global warming condition.
Abstract: A modified Genetic Algorithm (GA) based optimal selection of parameters for Automatic Generation Control (AGC) of multi-area electric energy systems is proposed in this paper. Simulations on multi-area reheat thermal system with and without consideration of nonlinearity like governor dead band followed by 1% step load perturbation is performed to exemplify the optimum parameter search. In this proposed method, a modified Genetic Algorithm is proposed where one point crossover with modification is employed. Positional dependency in respect of crossing site helps to maintain diversity of search point as well as exploitation of already known optimum value. This makes a trade-off between exploration and exploitation of search space to find global optimum in less number of generations. The proposed GA along with decomposition technique as developed has been used to obtain the optimum megawatt frequency control of multi-area electric energy systems. Time-domain simulations are conducted with trapezoidal integration along with decomposition technique. The superiority of the proposed method over existing one is verified from simulations and comparisons.
Abstract: We present new finite element methods for Helmholtz and Maxwell equations on general three-dimensional polyhedral meshes, based on domain decomposition with boundary elements on the surfaces of the polyhedral volume elements. The methods use the lowest-order polynomial spaces and produce sparse, symmetric linear systems despite the use of boundary elements. Moreover, piecewise constant coefficients are admissible. The resulting approximation on the element surfaces can be extended throughout the domain via representation formulas. Numerical experiments confirm that the convergence behavior on tetrahedral meshes is comparable to that of standard finite element methods, and equally good performance is attained on more general meshes.
Abstract: A new estimator for evolutionary spectrum (ES) based
on short time Fourier transform (STFT) and modified group delay
function (MGDF) by signal decomposition (SD) is proposed. The
STFT due to its built-in averaging, suppresses the cross terms and the
MGDF preserves the frequency resolution of the rectangular window
with the reduction in the Gibbs ripple. The present work overcomes
the magnitude distortion observed in multi-component non-stationary
signals with STFT and MGDF estimation of ES using SD. The SD is
achieved either through discrete cosine transform based harmonic
wavelet transform (DCTHWT) or perfect reconstruction filter banks
(PRFB). The MGDF also improves the signal to noise ratio by
removing associated noise. The performance of the present method is
illustrated for cross chirp and frequency shift keying (FSK) signals,
which indicates that its performance is better than STFT-MGDF
(STFT-GD) alone. Further its noise immunity is better than STFT.
The SD based methods, however cannot bring out the frequency
transition path from band to band clearly, as there will be gap in the
contour plot at the transition. The PRFB based STFT-SD shows good
performance than DCTHWT decomposition method for STFT-GD.
Abstract: High purity hydrogen and the valuable by-product of carbon nanotubes (CNTs) can be produced by the methane catalytic decomposition. The methane conversion and the performance of CNTs were determined by the choices of catalysts and the condition of decomposition reaction. In this paper, Ni/MgO and Ni/O-D (oxidized diamond) catalysts were prepared by wetness impregnation method. The effects of reaction temperature and space velocity of methane on the methane conversion were investigated in a fixed-bed. The surface area, structure and micrography were characterized with BET, XPS, SEM, EDS technology. The results showed that the conversion of methane was above 8% within 150 min (T=500) for 33Ni/O-D catalyst and higher than 25% within 120 min (T=650) for 41Ni/MgO catalyst. The initial conversion increased with the increasing temperature of the decomposition reaction, but their catalytic activities decreased rapidly while at too higher temperature. To decrease the space velocity of methane was propitious to promote the methane conversion, but not favor of the hydrogen yields. The appearance of carbon resulted from the methane decomposition lied on the support type and the condition of catalytic reaction. It presented as fiber shape on the surface of Ni/O-D at the relatively lower temperature such as 500 and 550, but as grain shape stacked on and overlayed on the surface of the metal nickel while at 650. The carbon fiber can form on the Ni/MgO surface at 650 and the diameter of the carbon fiber increased with the decreasing space velocity.
Abstract: In this paper, a new adaptive Fourier decomposition
(AFD) based time-frequency speech analysis approach is proposed.
Given the fact that the fundamental frequency of speech signals often
undergo fluctuation, the classical short-time Fourier transform (STFT)
based spectrogram analysis suffers from the difficulty of window size
selection. AFD is a newly developed signal decomposition theory. It is
designed to deal with time-varying non-stationary signals. Its
outstanding characteristic is to provide instantaneous frequency for
each decomposed component, so the time-frequency analysis becomes
easier. Experiments are conducted based on the sample sentence in
TIMIT Acoustic-Phonetic Continuous Speech Corpus. The results
show that the AFD based time-frequency distribution outperforms the
STFT based one.
Abstract: This paper presents a practical scheme that can be used for allocating the transmission loss to generators and loads. In this scheme first the share of a generator or load on the current through a branch is determined using Z-bus modified matrix. Then the current components are decomposed and the branch loss allocation is obtained. A motivation of proposed scheme is to improve the results of Z-bus method and to reach more fair allocation. The proposed scheme has been implemented and tested on several networks. To achieve practical and applicable results, the proposed scheme is simulated and compared on the transmission network (400kv) of Khorasan region in Iran and the 14-bus standard IEEE network. The results show that the proposed scheme is comprehensive and fair to allocating the energy losses of a power market to its participants.
Abstract: Wavelet transforms are multiresolution
decompositions that can be used to analyze signals and images.
Image compression is one of major applications of wavelet
transforms in image processing. It is considered as one of the most
powerful methods that provides a high compression ratio. However,
its implementation is very time-consuming. At the other hand,
parallel computing technologies are an efficient method for image
compression using wavelets. In this paper, we propose a parallel
wavelet compression algorithm based on quadtrees. We implement
the algorithm using MatlabMPI (a parallel, message passing version
of Matlab), and compute its isoefficiency function, and show that it is
scalable. Our experimental results confirm the efficiency of the
algorithm also.
Abstract: In this study, two new classes of generalized homeomorphisms are introduced and shown that one of these classes has a group structure. Moreover, some properties of these two homeomorphisms are obtained.
Abstract: An algorithm for learning an overcomplete dictionary
using a Cauchy mixture model for sparse decomposition of an underdetermined
mixing system is introduced. The mixture density
function is derived from a ratio sample of the observed mixture
signals where 1) there are at least two but not necessarily more
mixture signals observed, 2) the source signals are statistically
independent and 3) the sources are sparse. The basis vectors of the
dictionary are learned via the optimization of the location parameters
of the Cauchy mixture components, which is shown to be more
accurate and robust than the conventional data mining methods
usually employed for this task. Using a well known sparse
decomposition algorithm, we extract three speech signals from two
mixtures based on the estimated dictionary. Further tests with
additive Gaussian noise are used to demonstrate the proposed
algorithm-s robustness to outliers.
Abstract: In this paper, image compression using hybrid vector
quantization scheme such as Multistage Vector Quantization
(MSVQ) and Pyramid Vector Quantization (PVQ) are introduced. A
combined MSVQ and PVQ are utilized to take advantages provided
by both of them. In the wavelet decomposition of the image, most of
the information often resides in the lowest frequency subband.
MSVQ is applied to significant low frequency coefficients. PVQ is
utilized to quantize the coefficients of other high frequency
subbands. The wavelet coefficients are derived using lifting scheme.
The main aim of the proposed scheme is to achieve high compression
ratio without much compromise in the image quality. The results are
compared with the existing image compression scheme using MSVQ.
Abstract: In this work, the primary compressive strength
components of human femur trabecular bone are qualitatively
assessed using image processing and wavelet analysis. The Primary
Compressive (PC) component in planar radiographic femur trabecular
images (N=50) is delineated by semi-automatic image processing
procedure. Auto threshold binarization algorithm is employed to
recognize the presence of mineralization in the digitized images. The
qualitative parameters such as apparent mineralization and total area
associated with the PC region are derived for normal and abnormal
images.The two-dimensional discrete wavelet transforms are utilized
to obtain appropriate features that quantify texture changes in medical
images .The normal and abnormal samples of the human femur are
comprehensively analyzed using Harr wavelet.The six statistical
parameters such as mean, median, mode, standard deviation, mean
absolute deviation and median absolute deviation are derived at level
4 decomposition for both approximation and horizontal wavelet
coefficients. The correlation coefficient of various wavelet derived
parameters with normal and abnormal for both approximated and
horizontal coefficients are estimated. It is seen that in almost all cases
the abnormal show higher degree of correlation than normals. Further
the parameters derived from approximation coefficient show more
correlation than those derived from the horizontal coefficients. The
parameters mean and median computed at the output of level 4 Harr
wavelet channel was found to be a useful predictor to delineate the
normal and the abnormal groups.
Abstract: In this paper, we propose a supervised method for
color image classification based on a multilevel sigmoidal neural
network (MSNN) model. In this method, images are classified into
five categories, i.e., “Car", “Building", “Mountain", “Farm" and
“Coast". This classification is performed without any segmentation
processes. To verify the learning capabilities of the proposed method,
we compare our MSNN model with the traditional Sigmoidal Neural
Network (SNN) model. Results of comparison have shown that the
MSNN model performs better than the traditional SNN model in the
context of training run time and classification rate. Both color
moments and multi-level wavelets decomposition technique are used
to extract features from images. The proposed method has been
tested on a variety of real and synthetic images.
Abstract: The Goursat partial differential equation arises in
linear and non linear partial differential equations with mixed
derivatives. This equation is a second order hyperbolic partial
differential equation which occurs in various fields of study such as
in engineering, physics, and applied mathematics. There are many
approaches that have been suggested to approximate the solution of
the Goursat partial differential equation. However, all of the
suggested methods traditionally focused on numerical differentiation
approaches including forward and central differences in deriving the
scheme. An innovation has been done in deriving the Goursat partial
differential equation scheme which involves numerical integration
techniques. In this paper we have developed a new scheme to solve
the Goursat partial differential equation based on the Adomian
decomposition (ADM) and associated with Boole-s integration rule to
approximate the integration terms. The new scheme can easily be
applied to many linear and non linear Goursat partial differential
equations and is capable to reduce the size of computational work.
The accuracy of the results reveals the advantage of this new scheme
over existing numerical method.
Abstract: Wavelets have provided the researchers with
significant positive results, by entering the texture defect detection domain. The weak point of wavelets is that they are one-dimensional
by nature so they are not efficient enough to describe and analyze two-dimensional functions. In this paper we present a new method to
detect the defect of texture images by using curvelet transform.
Simulation results of the proposed method on a set of standard
texture images confirm its correctness. Comparing the obtained results indicates the ability of curvelet transform in describing
discontinuity in two-dimensional functions compared to wavelet
transform
Abstract: The main objective of this work is to provide a fault detection and isolation based on Markov parameters for residual generation and a neural network for fault classification. The diagnostic approach is accomplished in two steps: In step 1, the system is identified using a series of input / output variables through an identification algorithm. In step 2, the fault is diagnosed comparing the Markov parameters of faulty and non faulty systems. The Artificial Neural Network is trained using predetermined faulty conditions serves to classify the unknown fault. In step 1, the identification is done by first formulating a Hankel matrix out of Input/ output variables and then decomposing the matrix via singular value decomposition technique. For identifying the system online sliding window approach is adopted wherein an open slit slides over a subset of 'n' input/output variables. The faults are introduced at arbitrary instances and the identification is carried out in online. Fault residues are extracted making a comparison of the first five Markov parameters of faulty and non faulty systems. The proposed diagnostic approach is illustrated on benchmark problems with encouraging results.