Abstract: Reverse engineering of full-genomic interaction networks based on compendia of expression data has been successfully applied for a number of model organisms. This study adapts these approaches for an important non-model organism: The major human fungal pathogen Candida albicans. During the infection process, the pathogen can adapt to a wide range of environmental niches and reversibly changes its growth form. Given the importance of these processes, it is important to know how they are regulated. This study presents a reverse engineering strategy able to infer fullgenomic interaction networks for C. albicans based on a linear regression, utilizing the sparseness criterion (LASSO). To overcome the limited amount of expression data and small number of known interactions, we utilize different prior-knowledge sources guiding the network inference to a knowledge driven solution. Since, no database of known interactions for C. albicans exists, we use a textmining system which utilizes full-text research papers to identify known regulatory interactions. By comparing with these known regulatory interactions, we find an optimal value for global modelling parameters weighting the influence of the sparseness criterion and the prior-knowledge. Furthermore, we show that soft integration of prior-knowledge additionally improves the performance. Finally, we compare the performance of our approach to state of the art network inference approaches.
Abstract: User-based Collaborative filtering (CF), one of the
most prevailing and efficient recommendation techniques, provides
personalized recommendations to users based on the opinions of other
users. Although the CF technique has been successfully applied in
various applications, it suffers from serious sparsity problems. The
cloud-model approach addresses the sparsity problems by
constructing the user-s global preference represented by a cloud
eigenvector. The user-based CF approach works well with dense
datasets while the cloud-model CF approach has a greater
performance when the dataset is sparse. In this paper, we present a
hybrid approach that integrates the predictions from both the
user-based CF and the cloud-model CF approaches. The experimental
results show that the proposed hybrid approach can ameliorate the
sparsity problem and provide an improved prediction quality.
Abstract: A proof of convergence of a new continuation algorithm for computing the Analytic SVD for a large sparse parameter– dependent matrix is given. The algorithm itself was developed and numerically tested in [5].
Abstract: In comparison to the original SVM, which involves a
quadratic programming task; LS–SVM simplifies the required
computation, but unfortunately the sparseness of standard SVM is
lost. Another problem is that LS-SVM is only optimal if the training
samples are corrupted by Gaussian noise. In Least Squares SVM
(LS–SVM), the nonlinear solution is obtained, by first mapping the
input vector to a high dimensional kernel space in a nonlinear
fashion, where the solution is calculated from a linear equation set. In
this paper a geometric view of the kernel space is introduced, which
enables us to develop a new formulation to achieve a sparse and
robust estimate.
Abstract: Sparse representation which can represent high dimensional
data effectively has been successfully used in computer vision
and pattern recognition problems. However, it doesn-t consider the
label information of data samples. To overcome this limitation,
we develop a novel dimensionality reduction algorithm namely
dscriminatively regularized sparse subspace learning(DR-SSL) in this
paper. The proposed DR-SSL algorithm can not only make use of
the sparse representation to model the data, but also can effective
employ the label information to guide the procedure of dimensionality
reduction. In addition,the presented algorithm can effectively deal
with the out-of-sample problem.The experiments on gene-expression
data sets show that the proposed algorithm is an effective tool for
dimensionality reduction and gene-expression data classification.
Abstract: In this paper, the implementation of a rule-based
intuitive reasoner is presented. The implementation included two
parts: the rule induction module and the intuitive reasoner. A large
weather database was acquired as the data source. Twelve weather
variables from those data were chosen as the “target variables"
whose values were predicted by the intuitive reasoner. A “complex"
situation was simulated by making only subsets of the data available
to the rule induction module. As a result, the rules induced were
based on incomplete information with variable levels of certainty.
The certainty level was modeled by a metric called "Strength of
Belief", which was assigned to each rule or datum as ancillary
information about the confidence in its accuracy. Two techniques
were employed to induce rules from the data subsets: decision tree
and multi-polynomial regression, respectively for the discrete and the
continuous type of target variables. The intuitive reasoner was tested
for its ability to use the induced rules to predict the classes of the
discrete target variables and the values of the continuous target
variables. The intuitive reasoner implemented two types of
reasoning: fast and broad where, by analogy to human thought, the
former corresponds to fast decision making and the latter to deeper
contemplation. . For reference, a weather data analysis approach
which had been applied on similar tasks was adopted to analyze the
complete database and create predictive models for the same 12
target variables. The values predicted by the intuitive reasoner and
the reference approach were compared with actual data. The intuitive
reasoner reached near-100% accuracy for two continuous target
variables. For the discrete target variables, the intuitive reasoner
predicted at least 70% as accurately as the reference reasoner. Since
the intuitive reasoner operated on rules derived from only about 10%
of the total data, it demonstrated the potential advantages in dealing
with sparse data sets as compared with conventional methods.
Abstract: This paper presents a new feature based dense stereo
matching algorithm to obtain the dense disparity map via dynamic
programming. After extraction of some proper features, we use some
matching constraints such as epipolar line, disparity limit, ordering
and limit of directional derivative of disparity as well. Also, a coarseto-
fine multiresolution strategy is used to decrease the search space
and therefore increase the accuracy and processing speed. The
proposed method links the detected feature points into the chains and
compares some of the feature points from different chains, to
increase the matching speed. We also employ color stereo matching
to increase the accuracy of the algorithm. Then after feature
matching, we use the dynamic programming to obtain the dense
disparity map. It differs from the classical DP methods in the stereo
vision, since it employs sparse disparity map obtained from the
feature based matching stage. The DP is also performed further on a
scan line, between any matched two feature points on that scan line.
Thus our algorithm is truly an optimization method. Our algorithm
offers a good trade off in terms of accuracy and computational
efficiency. Regarding the results of our experiments, the proposed
algorithm increases the accuracy from 20 to 70%, and reduces the
running time of the algorithm almost 70%.
Abstract: The advantage of solving the complex nonlinear
problems by utilizing fuzzy logic methodologies is that the
experience or expert-s knowledge described as a fuzzy rule base can
be directly embedded into the systems for dealing with the problems.
The current limitation of appropriate and automated designing of
fuzzy controllers are focused in this paper. The structure discovery
and parameter adjustment of the Branched T-S fuzzy model is
addressed by a hybrid technique of type constrained sparse tree
algorithms. The simulation result for different system model is
evaluated and the identification error is observed to be minimum.
Abstract: Delay-Tolerant Networks (DTNs) are sparse, wireless
networks where disconnections are common due to host mobility and
low node density. The Message Ferrying (MF) scheme is a mobilityassisted
paradigm to improve connectivity in DTN-like networks. A
ferry or message ferry is a special node in the network which has
a per-determined route in the deployed area and relays messages
between mobile hosts (MHs) which are intermittently connected.
Increased contact opportunities among mobile hosts and the ferry
improve the performance of the network, both in terms of message
delivery ratio and average end-end delay. However, due to the inherent
mobility of mobile hosts and pre-determined periodicity of the
message ferry, mobile hosts may often -miss- contact opportunities
with a ferry. In this paper, we propose the combination of stationary
ferry access points (FAPs) with MF routing to increase contact
opportunities between mobile hosts and the MF and consequently
improve the performance of the DTN. We also propose several
placement models for deploying FAPs on MF routes. We evaluate the
performance of the FAP placement models through comprehensive
simulation. Our findings show that FAPs do improve the performance
of MF-assisted DTNs and symmetric placement of FAPs outperforms
other placement strategies.
Abstract: In the present paper, we propose numerical methods for solving the Stein equation AXC - X - D = 0 where the matrix A is large and sparse. Such problems appear in discrete-time control problems, filtering and image restoration. We consider the case where the matrix D is of full rank and the case where D is factored as a product of two matrices. The proposed methods are Krylov subspace methods based on the block Arnoldi algorithm. We give theoretical results and we report some numerical experiments.
Abstract: Restarted GMRES methods augmented with approximate eigenvectors are widely used for solving large sparse linear systems. Recently a new scheme of augmenting with error approximations is proposed. The main aim of this paper is to develop a restarted GMRES method augmented with the combination of harmonic Ritz vectors and error approximations. We demonstrate that the resulted combination method can gain the advantages of two approaches: (i) effectively deflate the small eigenvalues in magnitude that may hamper the convergence of the method and (ii) partially recover the global optimality lost due to restarting. The effectiveness and efficiency of the new method are demonstrated through various numerical examples.
Abstract: Text categorization is the problem of classifying text
documents into a set of predefined classes. After a preprocessing
step, the documents are typically represented as large sparse vectors.
When training classifiers on large collections of documents, both the
time and memory restrictions can be quite prohibitive. This justifies
the application of feature selection methods to reduce the
dimensionality of the document-representation vector. In this paper,
three feature selection methods are evaluated: Random Selection,
Information Gain (IG) and Support Vector Machine feature selection
(called SVM_FS). We show that the best results were obtained with
SVM_FS method for a relatively small dimension of the feature
vector. Also we present a novel method to better correlate SVM
kernel-s parameters (Polynomial or Gaussian kernel).
Abstract: In-core memory requirement is a bottleneck in solving
large three dimensional Navier-Stokes finite element problem
formulations using sparse direct solvers. Out-of-core solution
strategy is a viable alternative to reduce the in-core memory
requirements while solving large scale problems. This study
evaluates the performance of various out-of-core sequential solvers
based on multifrontal or supernodal techniques in the context of
finite element formulations for three dimensional problems on a
Windows platform. Here three different solvers, HSL_MA78,
MUMPS and PARDISO are compared. The performance of these
solvers is evaluated on a 64-bit machine with 16GB RAM for finite
element formulation of flow through a rectangular channel. It is
observed that using out-of-core PARDISO solver, relatively large
problems can be solved. The implementation of Newton and
modified Newton's iteration is also discussed.
Abstract: HSDPA is a new feature which is introduced in
Release-5 specifications of the 3GPP WCDMA/UTRA standard to
realize higher speed data rate together with lower round-trip times.
Moreover, the HSDPA concept offers outstanding improvement of
packet throughput and also significantly reduces the packet call
transfer delay as compared to Release -99 DSCH. Till now the
HSDPA system uses turbo coding which is the best coding technique
to achieve the Shannon limit. However, the main drawbacks of turbo
coding are high decoding complexity and high latency which makes
it unsuitable for some applications like satellite communications,
since the transmission distance itself introduces latency due to
limited speed of light. Hence in this paper it is proposed to use LDPC
coding in place of Turbo coding for HSDPA system which decreases
the latency and decoding complexity. But LDPC coding increases the
Encoding complexity. Though the complexity of transmitter
increases at NodeB, the End user is at an advantage in terms of
receiver complexity and Bit- error rate. In this paper LDPC Encoder
is implemented using “sparse parity check matrix" H to generate a
codeword at Encoder and “Belief Propagation algorithm "for LDPC
decoding .Simulation results shows that in LDPC coding the BER
suddenly drops as the number of iterations increase with a small
increase in Eb/No. Which is not possible in Turbo coding. Also same
BER was achieved using less number of iterations and hence the
latency and receiver complexity has decreased for LDPC coding.
HSDPA increases the downlink data rate within a cell to a theoretical
maximum of 14Mbps, with 2Mbps on the uplink. The changes that
HSDPA enables includes better quality, more reliable and more
robust data services. In other words, while realistic data rates are
only a few Mbps, the actual quality and number of users achieved
will improve significantly.
Abstract: In this paper, we introduce a new method for elliptical
object identification. The proposed method adopts a hybrid scheme
which consists of Eigen values of covariance matrices, Circular
Hough transform and Bresenham-s raster scan algorithms. In this
approach we use the fact that the large Eigen values and small Eigen
values of covariance matrices are associated with the major and minor
axial lengths of the ellipse. The centre location of the ellipse can be
identified using circular Hough transform (CHT). Sparse matrix
technique is used to perform CHT. Since sparse matrices squeeze zero
elements and contain a small number of nonzero elements they
provide an advantage of matrix storage space and computational time.
Neighborhood suppression scheme is used to find the valid Hough
peaks. The accurate position of circumference pixels is identified
using raster scan algorithm which uses the geometrical symmetry
property. This method does not require the evaluation of tangents or
curvature of edge contours, which are generally very sensitive to
noise working conditions. The proposed method has the advantages of
small storage, high speed and accuracy in identifying the feature. The
new method has been tested on both synthetic and real images.
Several experiments have been conducted on various images with
considerable background noise to reveal the efficacy and robustness.
Experimental results about the accuracy of the proposed method,
comparisons with Hough transform and its variants and other
tangential based methods are reported.
Abstract: In this paper we present a new method for coin
identification. The proposed method adopts a hybrid scheme using
Eigenvalues of covariance matrix, Circular Hough Transform (CHT)
and Bresenham-s circle algorithm. The statistical and geometrical
properties of the small and large Eigenvalues of the covariance
matrix of a set of edge pixels over a connected region of support are
explored for the purpose of circular object detection. Sparse matrix
technique is used to perform CHT. Since sparse matrices squeeze
zero elements and contain only a small number of non-zero elements,
they provide an advantage of matrix storage space and computational
time. Neighborhood suppression scheme is used to find the valid
Hough peaks. The accurate position of the circumference pixels is
identified using Raster scan algorithm which uses geometrical
symmetry property. After finding circular objects, the proposed
method uses the texture on the surface of the coins called texton,
which are unique properties of coins, refers to the fundamental micro
structure in generic natural images. This method has been tested on
several real world images including coin and non-coin images. The
performance is also evaluated based on the noise withstanding
capability.
Abstract: Sparse representation has long been studied and several
dictionary learning methods have been proposed. The dictionary
learning methods are widely used because they are adaptive. In this
paper, a new dictionary learning method for audio is proposed. Signals
are at first decomposed into different degrees of Intrinsic Mode
Functions (IMF) using Empirical Mode Decomposition (EMD)
technique. Then these IMFs form a learned dictionary. To reduce the
size of the dictionary, the K-means method is applied to the dictionary
to generate a K-EMD dictionary. Compared to K-SVD algorithm, the
K-EMD dictionary decomposes audio signals into structured
components, thus the sparsity of the representation is increased by
34.4% and the SNR of the recovered audio signals is increased by
20.9%.
Abstract: The problem addressed herein is the efficient management of the Grid/Cluster intense computation involved, when the preconditioned Bi-CGSTAB Krylov method is employed for the iterative solution of the large and sparse linear system arising from the discretization of the Modified Helmholtz-Dirichlet problem by the Hermite Collocation method. Taking advantage of the Collocation ma-trix's red-black ordered structure we organize efficiently the whole computation and map it on a pipeline architecture with master-slave communication. Implementation, through MPI programming tools, is realized on a SUN V240 cluster, inter-connected through a 100Mbps and 1Gbps ethernet network,and its performance is presented by speedup measurements included.
Abstract: Although Face detection is not a recent activity in the
field of image processing, it is still an open area for research. The
greatest step in this field is the work reported by Viola and its recent
analogous is Huang et al. Both of them use similar features and also
similar training process. The former is just for detecting upright
faces, but the latter can detect multi-view faces in still grayscale
images using new features called 'sparse feature'. Finding these
features is very time consuming and inefficient by proposed methods.
Here, we propose a new approach for finding sparse features using a
genetic algorithm system. This method requires less computational
cost and gets more effective features in learning process for face
detection that causes more accuracy.
Abstract: Text categorization is the problem of classifying text documents into a set of predefined classes. After a preprocessing step the documents are typically represented as large sparse vectors. When training classifiers on large collections of documents, both the time and memory restrictions can be quite prohibitive. This justifies the application of features selection methods to reduce the dimensionality of the document-representation vector. Four feature selection methods are evaluated: Random Selection, Information Gain (IG), Support Vector Machine (called SVM_FS) and Genetic Algorithm with SVM (GA_FS). We showed that the best results were obtained with SVM_FS and GA_FS methods for a relatively small dimension of the features vector comparative with the IG method that involves longer vectors, for quite similar classification accuracies. Also we present a novel method to better correlate SVM kernel-s parameters (Polynomial or Gaussian kernel).