Abstract: The mitigation of crop loss due to damaging freezes
requires accurate air temperature prediction models. Previous work
established that the Ward-style artificial neural network (ANN) is a
suitable tool for developing such models. The current research
focused on developing ANN models with reduced average prediction
error by increasing the number of distinct observations used in
training, adding additional input terms that describe the date of an
observation, increasing the duration of prior weather data included in
each observation, and reexamining the number of hidden nodes used
in the network. Models were created to predict air temperature at
hourly intervals from one to 12 hours ahead. Each ANN model,
consisting of a network architecture and set of associated parameters,
was evaluated by instantiating and training 30 networks and
calculating the mean absolute error (MAE) of the resulting networks
for some set of input patterns. The inclusion of seasonal input terms,
up to 24 hours of prior weather information, and a larger number of
processing nodes were some of the improvements that reduced
average prediction error compared to previous research across all
horizons. For example, the four-hour MAE of 1.40°C was 0.20°C, or
12.5%, less than the previous model. Prediction MAEs eight and 12
hours ahead improved by 0.17°C and 0.16°C, respectively,
improvements of 7.4% and 5.9% over the existing model at these
horizons. Networks instantiating the same model but with different
initial random weights often led to different prediction errors. These
results strongly suggest that ANN model developers should consider
instantiating and training multiple networks with different initial
weights to establish preferred model parameters.
Abstract: The automatic discrimination of seismic signals is an important practical goal for earth-science observatories due to the large amount of information that they receive continuously. An essential discrimination task is to allocate the incoming signal to a group associated with the kind of physical phenomena producing it. In this paper, two classes of seismic signals recorded routinely in geophysical laboratory of the National Center for Scientific and Technical Research in Morocco are considered. They correspond to signals associated to local earthquakes and chemical explosions. The approach adopted for the development of an automatic discrimination system is a modular system composed by three blocs: 1) Representation, 2) Dimensionality reduction and 3) Classification. The originality of our work consists in the use of a new wavelet called "modified Mexican hat wavelet" in the representation stage. For the dimensionality reduction, we propose a new algorithm based on the random projection and the principal component analysis.
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 aims to provide an approach to predict the
performance of the product produced after multi-stages of
manufacturing processes, as well as the assembly. Such approach
aims to control and subsequently identify the relationship between
the process inputs and outputs so that a process engineer can more
accurately predict how the process output shall perform based on the
system inputs. The approach is guided by a six-sigma methodology to
obtain improved performance.
In this paper a case study of the manufacture of a hermetic
reciprocating compressor is presented. The application of artificial
neural networks (ANNs) technique is introduced to improve
performance prediction within this manufacturing environment. The
results demonstrate that the approach predicts accurately and
effectively.
Abstract: The Emergency Department of a medical center in
Taiwan cooperated to conduct the research. A predictive model of
triage system is contracted from the contract procedure, selection of
parameters to sample screening. 2,000 pieces of data needed for the
patients is chosen randomly by the computer. After three
categorizations of data mining (Multi-group Discriminant Analysis,
Multinomial Logistic Regression, Back-propagation Neural
Networks), it is found that Back-propagation Neural Networks can
best distinguish the patients- extent of emergency, and the accuracy
rate can reach to as high as 95.1%. The Back-propagation Neural
Networks that has the highest accuracy rate is simulated into the triage
acuity expert system in this research. Data mining applied to the
predictive model of the triage acuity expert system can be updated
regularly for both the improvement of the system and for education
training, and will not be affected by subjective factors.
Abstract: Fundamental sensor-motor couplings form the backbone
of most mobile robot control tasks, and often need to be implemented
fast, efficiently and nevertheless reliably. Machine learning
techniques are therefore often used to obtain the desired sensor-motor
competences.
In this paper we present an alternative to established machine
learning methods such as artificial neural networks, that is very fast,
easy to implement, and has the distinct advantage that it generates
transparent, analysable sensor-motor couplings: system identification
through nonlinear polynomial mapping.
This work, which is part of the RobotMODIC project at the
universities of Essex and Sheffield, aims to develop a theoretical understanding
of the interaction between the robot and its environment.
One of the purposes of this research is to enable the principled design
of robot control programs.
As a first step towards this aim we model the behaviour of the
robot, as this emerges from its interaction with the environment, with
the NARMAX modelling method (Nonlinear, Auto-Regressive, Moving
Average models with eXogenous inputs). This method produces
explicit polynomial functions that can be subsequently analysed using
established mathematical methods.
In this paper we demonstrate the fidelity of the obtained NARMAX
models in the challenging task of robot route learning; we present a
set of experiments in which a Magellan Pro mobile robot was taught
to follow four different routes, always using the same mechanism to
obtain the required control law.
Abstract: Yeast cells live in a constantly changing environment that requires the continuous adaptation of their genomic program in order to sustain their homeostasis, survive and proliferate. Due to the advancement of high throughput technologies, there is currently a large amount of data such as gene expression, gene deletion and protein-protein interactions for S. Cerevisiae under various environmental conditions. Mining these datasets requires efficient computational methods capable of integrating different types of data, identifying inter-relations between different components and inferring functional groups or 'modules' that shape intracellular processes. This study uses computational methods to delineate some of the mechanisms used by yeast cells to respond to environmental changes. The GRAM algorithm is first used to integrate gene expression data and ChIP-chip data in order to find modules of coexpressed and co-regulated genes as well as the transcription factors (TFs) that regulate these modules. Since transcription factors are themselves transcriptionally regulated, a three-layer regulatory cascade consisting of the TF-regulators, the TFs and the regulated modules is subsequently considered. This three-layer cascade is then modeled quantitatively using artificial neural networks (ANNs) where the input layer corresponds to the expression of the up-stream transcription factors (TF-regulators) and the output layer corresponds to the expression of genes within each module. This work shows that (a) the expression of at least 33 genes over time and for different stress conditions is well predicted by the expression of the top layer transcription factors, including cases in which the effect of up-stream regulators is shifted in time and (b) identifies at least 6 novel regulatory interactions that were not previously associated with stress-induced changes in gene expression. These findings suggest that the combination of gene expression and protein-DNA interaction data with artificial neural networks can successfully model biological pathways and capture quantitative dependencies between distant regulators and downstream genes.
Abstract: In this paper a new fast simplification method is presented. Such method realizes Karnough map with large number of variables. In order to accelerate the operation of the proposed method, a new approach for fast detection of group of ones is presented. Such approach implemented in the frequency domain. The search operation relies on performing cross correlation in the frequency domain rather than time one. It is proved mathematically and practically that the number of computation steps required for the presented method is less than that needed by conventional cross correlation. Simulation results using MATLAB confirm the theoretical computations. Furthermore, a powerful solution for realization of complex functions is given. The simplified functions are implemented by using a new desigen for neural networks. Neural networks are used because they are fault tolerance and as a result they can recognize signals even with noise or distortion. This is very useful for logic functions used in data and computer communications. Moreover, the implemented functions are realized with minimum amount of components. This is done by using modular neural nets (MNNs) that divide the input space into several homogenous regions. Such approach is applied to implement XOR function, 16 logic functions on one bit level, and 2-bit digital multiplier. Compared to previous non- modular designs, a clear reduction in the order of computations and hardware requirements is achieved.
Abstract: In the artificial intelligence field, knowledge
representation and reasoning are important areas for intelligent
systems, especially knowledge base systems and expert systems.
Knowledge representation Methods has an important role in
designing the systems. There have been many models for knowledge
such as semantic networks, conceptual graphs, and neural networks.
These models are useful tools to design intelligent systems. However,
they are not suitable to represent knowledge in the domains of reality
applications. In this paper, new models for knowledge representation
called computational networks will be presented. They have been
used in designing some knowledge base systems in education for
solving problems such as the system that supports studying
knowledge and solving analytic geometry problems, the program for
studying and solving problems in Plane Geometry, the program for
solving problems about alternating current in physics.
Abstract: This paper proposes an effective adaptation learning
algorithm based on artificial neural networks for speed control of an
induction motor assumed to operate in a high-performance drives
environment. The structure scheme consists of a neural network
controller and an algorithm for changing the NN weights in order that
the motor speed can accurately track of the reference command. This
paper also makes uses a very realistic and practical scheme to
estimate and adaptively learn the noise content in the speed load
torque characteristic of the motor. The availability of the proposed
controller is verified by through a laboratory implementation and
under computation simulations with Matlab-software. The process is
also tested for the tracking property using different types of reference
signals. The performance and robustness of the proposed control
scheme have evaluated under a variety of operating conditions of the
induction motor drives. The obtained results demonstrate the
effectiveness of the proposed control scheme system performances,
both in steady state error in speed and dynamic conditions, was found
to be excellent and those is not overshoot.
Abstract: The System Identification problem looks for a
suitably parameterized model, representing a given process. The
parameters of the model are adjusted to optimize a performance
function based on error between the given process output and
identified process output. The linear system identification field is
well established with many classical approaches whereas most of
those methods cannot be applied for nonlinear systems. The problem
becomes tougher if the system is completely unknown with only the
output time series is available. It has been reported that the
capability of Artificial Neural Network to approximate all linear and
nonlinear input-output maps makes it predominantly suitable for the
identification of nonlinear systems, where only the output time series
is available. [1][2][4][5]. The work reported here is an attempt to
implement few of the well known algorithms in the context of
modeling of nonlinear systems, and to make a performance
comparison to establish the relative merits and demerits.
Abstract: In this research study, an intelligent detection system
to support medical diagnosis and detection of abnormal lesions by
processing endoscopic images is presented. The images used in this
study have been obtained using the M2A Swallowable Imaging
Capsule - a patented, video color-imaging disposable capsule.
Schemes have been developed to extract texture features from the
fuzzy texture spectra in the chromatic and achromatic domains for a
selected region of interest from each color component histogram of
endoscopic images. The implementation of an advanced fuzzy
inference neural network which combines fuzzy systems and
artificial neural networks and the concept of fusion of multiple
classifiers dedicated to specific feature parameters have been also
adopted in this paper. The achieved high detection accuracy of the
proposed system has provided thus an indication that such intelligent
schemes could be used as a supplementary diagnostic tool in
endoscopy.
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: This paper proposes new hybrid approaches for face
recognition. Gabor wavelets representation of face images is an
effective approach for both facial action recognition and face
identification. Perform dimensionality reduction and linear
discriminate analysis on the down sampled Gabor wavelet faces can
increase the discriminate ability. Nearest feature space is extended to
various similarity measures. In our experiments, proposed Gabor
wavelet faces combined with extended neural net feature space
classifier shows very good performance, which can achieve 93 %
maximum correct recognition rate on ORL data set without any preprocessing
step.
Abstract: Minimization methods for training feed-forward networks with Backpropagation are compared. Feedforward network training is a special case of functional minimization, where no explicit model of the data is assumed. Therefore due to the high dimensionality of the data, linearization of the training problem through use of orthogonal basis functions is not desirable. The focus is functional minimization on any basis. A number of methods based on local gradient and Hessian matrices are discussed. Modifications of many methods of first and second order training methods are considered. Using share rates data, experimentally it is proved that Conjugate gradient and Quasi Newton?s methods outperformed the Gradient Descent methods. In case of the Levenberg-Marquardt algorithm is of special interest in financial forecasting.
Abstract: Movable power sources of proton exchange
membrane fuel cells (PEMFC) are the important research done in the
current fuel cells (FC) field. The PEMFC system control influences
the cell performance greatly and it is a control system for industrial
complex problems, due to the imprecision, uncertainty and partial
truth and intrinsic nonlinear characteristics of PEMFCs. In this paper
an adaptive PI control strategy using neural network adaptive Morlet
wavelet for control is proposed. It is based on a single layer feed
forward neural networks with hidden nodes of adaptive morlet
wavelet functions controller and an infinite impulse response (IIR)
recurrent structure. The IIR is combined by cascading to the network
to provide double local structure resulting in improving speed of
learning. The proposed method is applied to a typical 1 KW PEMFC
system and the results show the proposed method has more accuracy
against to MLP (Multi Layer Perceptron) method.
Abstract: Although many researchers have studied the flow
hydraulics in compound channels, there are still many complicated problems in determination of their flow rating curves. Many different
methods have been presented for these channels but extending them
for all types of compound channels with different geometrical and
hydraulic conditions is certainly difficult. In this study, by aid of nearly 400 laboratory and field data sets of geometry and flow rating
curves from 30 different straight compound sections and using artificial neural networks (ANNs), flow discharge in compound channels was estimated. 13 dimensionless input variables including relative depth, relative roughness, relative width, aspect ratio, bed
slope, main channel side slopes, flood plains side slopes and berm
inclination and one output variable (flow discharge), have been used
in ANNs. Comparison of ANNs model and traditional method
(divided channel method-DCM) shows high accuracy of ANNs model results. The results of Sensitivity analysis showed that the relative depth with 47.6 percent contribution, is the most effective input parameter for flow discharge prediction. Relative width and
relative roughness have 19.3 and 12.2 percent of importance, respectively. On the other hand, shape parameter, main channel and
flood plains side slopes with 2.1, 3.8 and 3.8 percent of contribution, have the least importance.
Abstract: Text categorization - the assignment of natural language documents to one or more predefined categories based on their semantic content - is an important component in many information organization and management tasks. Performance of neural networks learning is known to be sensitive to the initial weights and architecture. This paper discusses the use multilayer neural network initialization with decision tree classifier for improving text categorization accuracy. An adaptation of the algorithm is proposed in which a decision tree from root node until a final leave is used for initialization of multilayer neural network. The experimental evaluation demonstrates this approach provides better classification accuracy with Reuters-21578 corpus, one of the standard benchmarks for text categorization tasks. We present results comparing the accuracy of this approach with multilayer neural network initialized with traditional random method and decision tree classifiers.
Abstract: Recent scientific investigations indicate that
multimodal biometrics overcome the technical limitations of
unimodal biometrics, making them ideally suited for everyday life
applications that require a reliable authentication system. However,
for a successful adoption of multimodal biometrics, such systems
would require large heterogeneous datasets with complex multimodal
fusion and privacy schemes spanning various distributed
environments. From experimental investigations of current
multimodal systems, this paper reports the various issues related to
speed, error-recovery and privacy that impede the diffusion of such
systems in real-life. This calls for a robust mechanism that caters to
the desired real-time performance, robust fusion schemes,
interoperability and adaptable privacy policies.
The main objective of this paper is to present a framework that
addresses the abovementioned issues by leveraging on the
heterogeneous resource sharing capacities of Grid services and the
efficient machine learning capabilities of artificial neural networks
(ANN). Hence, this paper proposes a Grid-based neural network
framework for adopting multimodal biometrics with the view of
overcoming the barriers of performance, privacy and risk issues that
are associated with shared heterogeneous multimodal data centres.
The framework combines the concept of Grid services for reliable
brokering and privacy policy management of shared biometric
resources along with a momentum back propagation ANN (MBPANN)
model of machine learning for efficient multimodal fusion and
authentication schemes. Real-life applications would be able to adopt
the proposed framework to cater to the varying business requirements
and user privacies for a successful diffusion of multimodal
biometrics in various day-to-day transactions.
Abstract: In contrast to existing methods which do not take into account multiconnectivity in a broad sense of this term, we develop mathematical models and highly effective combination (BIEM and FDM) numerical methods of calculation of stationary and cvazistationary temperature field of a profile part of a blade with convective cooling (from the point of view of realization on PC). The theoretical substantiation of these methods is proved by appropriate theorems. For it, converging quadrature processes have been developed and the estimations of errors in the terms of A.Ziqmound continuity modules have been received. For visualization of profiles are used: the method of the least squares with automatic conjecture, device spline, smooth replenishment and neural nets. Boundary conditions of heat exchange are determined from the solution of the corresponding integral equations and empirical relationships. The reliability of designed methods is proved by calculation and experimental investigations heat and hydraulic characteristics of the gas turbine 1st stage nozzle blade