Abstract: The paper is a comparative study of two classical vari-ants of parallel projection methods for solving the convex feasibility problem with their equivalents that involve variable weights in the construction of the solutions. We used a graphical representation of these methods for inpainting a convex area of an image in order to investigate their effectiveness in image reconstruction applications. We also presented a numerical analysis of the convergence of these four algorithms in terms of the average number of steps and execution time, in classical CPU and, alternativaly, in parallel GPU implementation.
Abstract: Over-parameterized neural networks have attracted a
great deal of attention in recent deep learning theory research,
as they challenge the classic perspective of over-fitting when
the model has excessive parameters and have gained empirical
success in various settings. While a number of theoretical works
have been presented to demystify properties of such models, the
convergence properties of such models are still far from being
thoroughly understood. In this work, we study the convergence
properties of training two-hidden-layer partially over-parameterized
fully connected networks with the Rectified Linear Unit activation via
gradient descent. To our knowledge, this is the first theoretical work
to understand convergence properties of deep over-parameterized
networks without the equally-wide-hidden-layer assumption and
other unrealistic assumptions. We provide a probabilistic lower bound
of the widths of hidden layers and proved linear convergence rate of
gradient descent. We also conducted experiments on synthetic and
real-world datasets to validate our theory.
Abstract: The present study deals with the finite element (FE) analysis of thermally-induced bistable plate using various plate elements. The quadrilateral plate elements include the 4-node conforming plate element based on the classical laminate plate theory (CLPT), the 4-node and 9-node Mindlin plate element based on the first-order shear deformation laminated plate theory (FSDT), and a displacement-based 4-node quadrilateral element (RDKQ-NL20). Using the von-Karman’s large deflection theory and the total Lagrangian (TL) approach, the nonlinear FE governing equations for plate under thermal load are derived. Convergence analysis for four elements is first conducted. These elements are then used to predict the stable shapes of thermally-induced bistable plate. Numerical test shows that the plate element based on FSDT, namely the 4-node and 9-node Mindlin, and the RDKQ-NL20 plate element can predict two stable cylindrical shapes while the 4-node conforming plate predicts a saddles shape. Comparing the simulation results with ABAQUS, the RDKQ-NL20 element shows the best accuracy among all the elements.
Abstract: The convergence rate of the least-mean-square (LMS)
algorithm deteriorates if the input signal to the filter is correlated.
In a system identification problem, this convergence rate can be
improved if the signal is white and/or if the system is sparse. We
recently proposed a sparse transform domain LMS-type algorithm
that uses a variable step-size for a sparse system identification.
The proposed algorithm provided high performance even if the
input signal is highly correlated. In this work, we investigate the
performance of the proposed TD-LMS algorithm for a large number
of filter tap which is also a critical issue for standard LMS algorithm.
Additionally, the optimum value of the most important parameter is
calculated for all experiments. Moreover, the convergence analysis
of the proposed algorithm is provided. The performance of the
proposed algorithm has been compared to different algorithms in a
sparse system identification setting of different sparsity levels and
different number of filter taps. Simulations have shown that the
proposed algorithm has prominent performance compared to the other
algorithms.
Abstract: Stochastic modeling concerns the use of probability
to model real-world situations in which uncertainty is present.
Therefore, the purpose of stochastic modeling is to estimate the
probability of outcomes within a forecast, i.e. to be able to predict
what conditions or decisions might happen under different situations.
In the present study, we present a model of a stochastic diffusion
process based on the bi-Weibull distribution function (its trend
is proportional to the bi-Weibull probability density function). In
general, the Weibull distribution has the ability to assume the
characteristics of many different types of distributions. This has
made it very popular among engineers and quality practitioners, who
have considered it the most commonly used distribution for studying
problems such as modeling reliability data, accelerated life testing,
and maintainability modeling and analysis. In this work, we start
by obtaining the probabilistic characteristics of this model, as the
explicit expression of the process, its trends, and its distribution by
transforming the diffusion process in a Wiener process as shown in
the Ricciaardi theorem. Then, we develop the statistical inference of
this model using the maximum likelihood methodology. Finally, we
analyse with simulated data the computational problems associated
with the parameters, an issue of great importance in its application to
real data with the use of the convergence analysis methods. Overall,
the use of a stochastic model reflects only a pragmatic decision on
the part of the modeler. According to the data that is available and
the universe of models known to the modeler, this model represents
the best currently available description of the phenomenon under
consideration.
Abstract: Model updating is an inverse eigenvalue problem which
concerns the modification of an existing but inaccurate model with
measured modal data. In this paper, an efficient gradient based
iterative method for updating the mass, damping and stiffness
matrices simultaneously using a few of complex measured modal
data is developed. Convergence analysis indicates that the iterative
solutions always converge to the unique minimum Frobenius norm
symmetric solution of the model updating problem by choosing a
special kind of initial matrices.
Abstract: In this paper, numerical solution of system of
Fredholm and Volterra integral equations by means of the Spline
collocation method is considered. This approximation reduces the
system of integral equations to an explicit system of algebraic
equations. The solution is collocated by cubic B-spline and the
integrand is approximated by the Newton-Cotes formula. The error
analysis of proposed numerical method is studied theoretically. The
results are compared with the results obtained by other methods to
illustrate the accuracy and the implementation of our method.
Abstract: Analysis of real life problems often results in linear
systems of equations for which solutions are sought. The method to
employ depends, to some extent, on the properties of the coefficient
matrix. It is not always feasible to solve linear systems of equations
by direct methods, as such the need to use an iterative method
becomes imperative. Before an iterative method can be employed
to solve a linear system of equations there must be a guaranty that
the process of solution will converge. This guaranty, which must
be determined apriori, involve the use of some criterion expressible
in terms of the entries of the coefficient matrix. It is, therefore,
logical that the convergence criterion should depend implicitly on the
algebraic structure of such a method. However, in deference to this
view is the practice of conducting convergence analysis for Gauss-
Seidel iteration on a criterion formulated based on the algebraic
structure of Jacobi iteration. To remedy this anomaly, the Gauss-
Seidel iteration was studied for its algebraic structure and contrary
to the usual assumption, it was discovered that some property of the
iteration matrix of Gauss-Seidel method is only diagonally dominant
in its first row while the other rows do not satisfy diagonal dominance.
With the aid of this structure we herein fashion out an improved
version of Gauss-Seidel iteration with the prospect of enhancing
convergence and robustness of the method. A numerical section is
included to demonstrate the validity of the theoretical results obtained
for the improved Gauss-Seidel method.
Abstract: Non-negative matrix factorization (NMF) is a useful computational method to find basis information of multivariate nonnegative data. A popular approach to solve the NMF problem is the multiplicative update (MU) algorithm. But, it has some defects. So the columnwisely alternating gradient (cAG) algorithm was proposed. In this paper, we analyze convergence of the cAG algorithm and show advantages over the MU algorithm. The stability of the equilibrium point is used to prove the convergence of the cAG algorithm. A classic model is used to obtain the equilibrium point and the invariant sets are constructed to guarantee the integrity of the stability. Finally, the convergence conditions of the cAG algorithm are obtained, which help reducing the evaluation time and is confirmed in the experiments. By using the same method, the MU algorithm has zero divisor and is convergent at zero has been verified. In addition, the convergence conditions of the MU algorithm at zero are similar to that of the cAG algorithm at non-zero. However, it is meaningless to discuss the convergence at zero, which is not always the result that we want for NMF. Thus, we theoretically illustrate the advantages of the cAG algorithm.
Abstract: This paper investigates a method for the state estimation of nonlinear systems described by a class of differential-algebraic equation (DAE) models using the extended Kalman filter. The method involves the use of a transformation from a DAE to ordinary differential equation (ODE). A relevant dynamic power system model using decoupled techniques will be proposed. The estimation technique consists of a state estimator based on the EKF technique as well as the local stability analysis. High performances are illustrated through a simulation study applied on IEEE 13 buses test system.
Abstract: We discuss the convergence property of the minimum residual (MINRES) method for the solution of complex shifted Hermitian system (αI + H)x = f. Our convergence analysis shows that the method has a faster convergence than that for real shifted Hermitian system (Re(α)I + H)x = f under the condition Re(α) + λmin(H) > 0, and a larger imaginary part of the shift α has a better convergence property. Numerical experiments show such convergence properties.
Abstract: In this paper, we develop quartic nonpolynomial
spline method for the numerical solution of third order two point
boundary value problems. It is shown that the new method gives
approximations, which are better than those produced by other spline
methods. Convergence analysis of the method is discussed through
standard procedures. Two numerical examples are given to illustrate
the applicability and efficiency of the novel method.
Abstract: In wireless communication system, a Decision Feedback Equalizer (DFE) to cancel the intersymbol interference (ISI) is required. In this paper, an exact convergence analysis of the (DFE) adapted by the Least Mean Square (LMS) algorithm during the training phase is derived by taking into account the finite alphabet context of data transmission. This allows us to determine the shortest training sequence that allows to reach a given Mean Square Error (MSE). With the intention of avoiding the problem of ill-convergence, the paper proposes an initialization strategy for the blind decision directed (DD) algorithm. This then yields a semi-blind DFE with high speed and good convergence.
Abstract: In this paper the behavior of the decision feedback
equalizers (DFEs) adapted by the decision-directed or the constant
modulus blind algorithms is presented. An analysis of the error
surface of the corresponding criterion cost functions is first
developed. With the intention of avoiding the ill-convergence of the
algorithm, the paper proposes to modify the shape of the cost
function error surface by using a soft decision instead of the hard
one. This was shown to reduce the influence of false decisions and to
smooth the undesirable minima. Modified algorithms using the soft
decision during a pseudo-training phase with an automatic switch to
the properly tracking phase are then derived. Computer simulations
show that these modified algorithms present better ability to avoid
local minima than conventional ones.
Abstract: According to the interaction of inflation and
unemployment, expectation of the rate of inflation in Croatia is
estimated. The interaction between inflation and unemployment is
shown by model based on three first-order differential i.e. difference
equations: Phillips relation, adaptive expectations equation and
monetary-policy equation. The resulting equation is second order
differential i.e. difference equation which describes the time path of
inflation. The data of the rate of inflation and the rate of
unemployment are used for parameters estimation. On the basis of
the estimated time paths, the stability and convergence analysis is
done for the rate of inflation.
Abstract: A kind of singularly perturbed boundary value problems is under consideration. In order to obtain its approximation, simple upwind difference discretization is applied. We use a moving mesh iterative algorithm based on equi-distributing of the arc-length function of the current computed piecewise linear solution. First, a maximum norm a posteriori error estimate on an arbitrary mesh is derived using a different method from the one carried out by Chen [Advances in Computational Mathematics, 24(1-4) (2006), 197-212.]. Then, basing on the properties of discrete Green-s function and the presented posteriori error estimate, we theoretically prove that the discrete solutions computed by the algorithm are first-order uniformly convergent with respect to the perturbation parameter ε.
Abstract: In this paper, we give the generalized alternating twostage method in which the inner iterations are accomplished by a generalized alternating method. And we present convergence results of the method for solving nonsingular linear systems when the coefficient matrix of the linear system is a monotone matrix or an H-matrix.
Abstract: In this paper we use quintic non-polynomial
spline functions to develop numerical methods for approximation
to the solution of a system of fourth-order boundaryvalue
problems associated with obstacle, unilateral and contact
problems. The convergence analysis of the methods has been
discussed and shown that the given approximations are better
than collocation and finite difference methods. Numerical
examples are presented to illustrate the applications of these
methods, and to compare the computed results with other
known methods.
Abstract: This paper presents the convergence analysis
of a prediction based blind equalizer for IIR channels.
Predictor parameters are estimated by using the recursive
least squares algorithm. It is shown that the prediction
error converges almost surely (a.s.) toward a scalar
multiple of the unknown input symbol sequence. It is
also proved that the convergence rate of the parameter
estimation error is of the same order as that in the iterated
logarithm law.