Abstract: Since the presentation of the backpropagation algorithm, a vast variety of improvements of the technique for training a feed forward neural networks have been proposed. This article focuses on two classes of acceleration techniques, one is known as Local Adaptive Techniques that are based on weightspecific only, such as the temporal behavior of the partial derivative of the current weight. The other, known as Dynamic Adaptation Methods, which dynamically adapts the momentum factors, α, and learning rate, η, with respect to the iteration number or gradient. Some of most popular learning algorithms are described. These techniques have been implemented and tested on several problems and measured in terms of gradient and error function evaluation, and percentage of success. Numerical evidence shows that these techniques improve the convergence of the Backpropagation algorithm.
Abstract: Clustering techniques have received attention in many areas including engineering, medicine, biology and data mining. The purpose of clustering is to group together data points, which are close to one another. The K-means algorithm is one of the most widely used techniques for clustering. However, K-means has two shortcomings: dependency on the initial state and convergence to local optima and global solutions of large problems cannot found with reasonable amount of computation effort. In order to overcome local optima problem lots of studies done in clustering. This paper is presented an efficient hybrid evolutionary optimization algorithm based on combining Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO), called PSO-ACO, for optimally clustering N object into K clusters. The new PSO-ACO algorithm is tested on several data sets, and its performance is compared with those of ACO, PSO and K-means clustering. The simulation results show that the proposed evolutionary optimization algorithm is robust and suitable for handing data clustering.
Abstract: An iterative algorithm is proposed and tested in Cournot Game models, which is based on the convergence of sequential best responses and the utilization of a genetic algorithm for determining each player-s best response to a given strategy profile of its opponents. An extra outer loop is used, to address the problem of finite accuracy, which is inherent in genetic algorithms, since the set of feasible values in such an algorithm is finite. The algorithm is tested in five Cournot models, three of which have convergent best replies sequence, one with divergent sequential best replies and one with “local NE traps"[14], where classical local search algorithms fail to identify the Nash Equilibrium. After a series of simulations, we conclude that the algorithm proposed converges to the Nash Equilibrium, with any level of accuracy needed, in all but the case where the sequential best replies process diverges.
Abstract: The purpose of this paper is to analyze the case of the
U.S. Pivot and to suggest an appropriate model including entry
strategies and success factors for QPS of Cable TV. The
telecommunication companies have been operating QPS including
IPTV service, which enables them to cross over broadcasting areas.
Due to this circumstance, the Cable TV operators are now concerned
and are planning to add QPS with the mobile service. Based on the
Porter's five forces model, an analytical framework has been proposed
to MVNO in Cable TV industry in the United States. As a result of this
study, MVNO in Cable TV industry has to have a clear killer
application with their sufficient contents. Subsequently, the direction
of the future Cable TV industry is proposed.
Abstract: Various intelligences and inspirations have been
adopted into the iterative searching process called as meta-heuristics.
They intelligently perform the exploration and exploitation in the
solution domain space aiming to efficiently seek near optimal
solutions. In this work, the bee algorithm, inspired by the natural
foraging behaviour of honey bees, was adapted to find the near
optimal solutions of the transportation management system, dynamic
multi-zone dispatching. This problem prepares for an uncertainty and
changing customers- demand. In striving to remain competitive,
transportation system should therefore be flexible in order to cope
with the changes of customers- demand in terms of in-bound and outbound
goods and technological innovations. To remain higher service
level but lower cost management via the minimal imbalance scenario,
the rearrangement penalty of the area, in each zone, including time
periods are also included. However, the performance of the algorithm
depends on the appropriate parameters- setting and need to be
determined and analysed before its implementation. BEE parameters
are determined through the linear constrained response surface
optimisation or LCRSOM and weighted centroid modified simplex
methods or WCMSM. Experimental results were analysed in terms
of best solutions found so far, mean and standard deviation on the
imbalance values including the convergence of the solutions
obtained. It was found that the results obtained from the LCRSOM
were better than those using the WCMSM. However, the average
execution time of experimental run using the LCRSOM was longer
than those using the WCMSM. Finally a recommendation of proper
level settings of BEE parameters for some selected problem sizes is
given as a guideline for future applications.
Abstract: In this paper, we consider an iteration process for
approximating common fixed points of two asymptotically quasinonexpansive
mappings and we prove some strong and weak convergence
theorems for such mappings in uniformly convex Banach
spaces.
Abstract: This paper presents and evaluates a new classification
method that aims to improve classifiers performances and speed up
their training process. The proposed approach, called labeled
classification, seeks to improve convergence of the BP (Back
propagation) algorithm through the addition of an extra feature
(labels) to all training examples. To classify every new example, tests
will be carried out each label. The simplicity of implementation is the
main advantage of this approach because no modifications are
required in the training algorithms. Therefore, it can be used with
others techniques of acceleration and stabilization. In this work, two
models of the labeled classification are proposed: the LMLP
(Labeled Multi Layered Perceptron) and the LNFC (Labeled Neuro
Fuzzy Classifier). These models are tested using Iris, wine, texture
and human thigh databases to evaluate their performances.
Abstract: This paper presents a novel genetic algorithm, termed
the Optimum Individual Monogenetic Algorithm (OIMGA) and
describes its hardware implementation. As the monogenetic strategy
retains only the optimum individual, the memory requirement is
dramatically reduced and no crossover circuitry is needed, thereby
ensuring the requisite silicon area is kept to a minimum.
Consequently, depending on application requirements, OIMGA
allows the investigation of solutions that warrant either larger GA
populations or individuals of greater length. The results given in this
paper demonstrate that both the performance of OIMGA and its
convergence time are superior to those of existing hardware GA
implementations. Local convergence is achieved in OIMGA by
retaining elite individuals, while population diversity is ensured by
continually searching for the best individuals in fresh regions of the
search space.
Abstract: Hypersonic flows around spatial vehicles during their
reentry phase in planetary atmospheres are characterized by intense
aerothermal phenomena. The aim of this work is to analyze high
temperature flows around an axisymmetric blunt body taking into
account chemical and vibrational non-equilibrium for air mixture
species. For this purpose, a finite volume methodology is employed
to determine the supersonic flow parameters around the axisymmetric
blunt body, especially at the stagnation point and along the wall of
spacecraft for several altitudes. This allows the capture shock wave
before a blunt body placed in supersonic free stream. The numerical
technique uses the Flux Vector Splitting method of Van Leer. Here,
adequate time stepping parameter, along with CFL coefficient and
mesh size level are selected to ensure numerical convergence, sought
with an order of 10-8
Abstract: There are two common types of operational research techniques, optimisation and metaheuristic methods. The latter may be defined as a sequential process that intelligently performs the exploration and exploitation adopted by natural intelligence and strong inspiration to form several iterative searches. An aim is to effectively determine near optimal solutions in a solution space. In this work, a type of metaheuristics called Ant Colonies Optimisation, ACO, inspired by a foraging behaviour of ants was adapted to find optimal solutions of eight non-linear continuous mathematical models. Under a consideration of a solution space in a specified region on each model, sub-solutions may contain global or multiple local optimum. Moreover, the algorithm has several common parameters; number of ants, moves, and iterations, which act as the algorithm-s driver. A series of computational experiments for initialising parameters were conducted through methods of Rigid Simplex, RS, and Modified Simplex, MSM. Experimental results were analysed in terms of the best so far solutions, mean and standard deviation. Finally, they stated a recommendation of proper level settings of ACO parameters for all eight functions. These parameter settings can be applied as a guideline for future uses of ACO. This is to promote an ease of use of ACO in real industrial processes. It was found that the results obtained from MSM were pretty similar to those gained from RS. However, if these results with noise standard deviations of 1 and 3 are compared, MSM will reach optimal solutions more efficiently than RS, in terms of speed of convergence.
Abstract: In this contribution, the use of a new genetic operator is proposed. The main advantage of using this operator is that it is able to assist the evolution procedure to converge faster towards the optimal solution of a problem. This new genetic operator is called ''intuition'' operator. Generally speaking, one can claim that this operator is a way to include any heuristic or any other local knowledge, concerning the problem, that cannot be embedded in the fitness function. Simulation results show that the use of this operator increases significantly the performance of the classic Genetic Algorithm by increasing the convergence speed of its population.
Abstract: An upwind difference approximation is used for a singularly perturbed problem in material science. Based on the discrete Green-s function theory, the error estimate in maximum norm is achieved, which is first-order uniformly convergent with respect to the perturbation parameter. The numerical experimental result is verified the valid of the theoretical analysis.
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: In this paper, we use a one-step iteration scheme to approximate common fixed points of two quasi-asymptotically nonexpansive mappings. We prove weak and strong convergence theorems in a uniformly convex Banach space. Our results generalize the corresponding results of Yao and Chen [15] to a wider class of mappings while extend those of Khan, Abbas and Khan [4] to an improved one-step iteration scheme without any condition and improve upon many others in the literature.
Abstract: In this paper, a pipelined version of genetic algorithm,
called PLGA, and a corresponding hardware platform are described.
The basic operations of conventional GA (CGA) are made pipelined
using an appropriate selection scheme. The selection operator, used
here, is stochastic in nature and is called SA-selection. This helps
maintaining the basic generational nature of the proposed pipelined
GA (PLGA). A number of benchmark problems are used to compare
the performances of conventional roulette-wheel selection and the
SA-selection. These include unimodal and multimodal functions with
dimensionality varying from very small to very large. It is seen that
the SA-selection scheme is giving comparable performances with
respect to the classical roulette-wheel selection scheme, for all the
instances, when quality of solutions and rate of convergence are considered.
The speedups obtained by PLGA for different benchmarks
are found to be significant. It is shown that a complete hardware
pipeline can be developed using the proposed scheme, if parallel
evaluation of the fitness expression is possible. In this connection
a low-cost but very fast hardware evaluation unit is described.
Results of simulation experiments show that in a pipelined hardware
environment, PLGA will be much faster than CGA. In terms of
efficiency, PLGA is found to outperform parallel GA (PGA) also.
Abstract: Discrete particle swarm optimization (DPSO) is a
powerful stochastic evolutionary algorithm that is used to solve the
large-scale, discrete and nonlinear optimization problems. However,
it has been observed that standard DPSO algorithm has premature
convergence when solving a complex optimization problem like
transmission expansion planning (TEP). To resolve this problem an
advanced discrete particle swarm optimization (ADPSO) is proposed
in this paper. The simulation result shows that optimization of lines
loading in transmission expansion planning with ADPSO is better
than DPSO from precision view point.
Abstract: Hearing impairment is the number one chronic
disability affecting many people in the world. Background noise is
particularly damaging to speech intelligibility for people with
hearing loss especially for sensorineural loss patients. Several
investigations on speech intelligibility have demonstrated
sensorineural loss patients need 5-15 dB higher SNR than the normal
hearing subjects. This paper describes Discrete Hartley Transform
Power Normalized Least Mean Square algorithm (DHT-LMS) to
improve the SNR and to reduce the convergence rate of the Least
Means Square (LMS) for sensorineural loss patients. The DHT
transforms n real numbers to n real numbers, and has the convenient
property of being its own inverse. It can be effectively used for noise
cancellation with less convergence time. The simulated result shows
the superior characteristics by improving the SNR at least 9 dB for
input SNR with zero dB and faster convergence rate (eigenvalue ratio
12) compare to time domain method and DFT-LMS.
Abstract: The purpose of this research is to develop and apply the
RSCMAC to enhance the dynamic accuracy of Global Positioning
System (GPS). GPS devices provide services of accurate positioning,
speed detection and highly precise time standard for over 98% area on
the earth. The overall operation of Global Positioning System includes
24 GPS satellites in space; signal transmission that includes 2
frequency carrier waves (Link 1 and Link 2) and 2 sets random
telegraphic codes (C/A code and P code), on-earth monitoring stations
or client GPS receivers. Only 4 satellites utilization, the client position
and its elevation can be detected rapidly. The more receivable
satellites, the more accurate position can be decoded. Currently, the
standard positioning accuracy of the simplified GPS receiver is greatly
increased, but due to affected by the error of satellite clock, the
troposphere delay and the ionosphere delay, current measurement
accuracy is in the level of 5~15m. In increasing the dynamic GPS
positioning accuracy, most researchers mainly use inertial navigation
system (INS) and installation of other sensors or maps for the
assistance. This research utilizes the RSCMAC advantages of fast
learning, learning convergence assurance, solving capability of
time-related dynamic system problems with the static positioning
calibration structure to improve and increase the GPS dynamic
accuracy. The increasing of GPS dynamic positioning accuracy can be
achieved by using RSCMAC system with GPS receivers collecting
dynamic error data for the error prediction and follows by using the
predicted error to correct the GPS dynamic positioning data. The
ultimate purpose of this research is to improve the dynamic positioning
error of cheap GPS receivers and the economic benefits will be
enhanced while the accuracy is increased.
Abstract: An alternative iterative computational procedure is
proposed for internal normal ball loads calculation in statically
loaded single-row, angular-contact ball bearings, subjected to a
known thrust load, which is applied in the inner ring at the geometric
bearing center line. An accurate method for curvature radii at
contacts with inner and outer raceways in the direction of the motion
is used. Numerical aspects of the iterative procedure are discussed.
Numerical examples results for a 218 angular-contact ball bearing
have been compared with those from the literature. Twenty figures
are presented showing the geometrical features, the behavior of the
convergence variables and the following parameters as functions of
the thrust load: normal ball loads, contact angle, distance between
curvature centers, and normal ball and axial deflections.
Abstract: In this paper, we present parallel alternating two-stage
methods for solving linear system Ax=b, where A is a symmetric
positive definite matrix. And we give some convergence results of
these methods for nonsingular linear system.