Abstract: The conjugate gradient optimization algorithm
usually used for nonlinear least squares is presented and is
combined with the modified back propagation algorithm yielding
a new fast training multilayer perceptron (MLP) algorithm
(CGFR/AG). The approaches presented in the paper consist of
three steps: (1) Modification on standard back propagation
algorithm by introducing gain variation term of the activation
function, (2) Calculating the gradient descent on error with
respect to the weights and gains values and (3) the determination
of the new search direction by exploiting the information
calculated by gradient descent in step (2) as well as the previous
search direction. The proposed method improved the training
efficiency of back propagation algorithm by adaptively modifying
the initial search direction. Performance of the proposed method
is demonstrated by comparing to the conjugate gradient algorithm
from neural network toolbox for the chosen benchmark. The
results show that the number of iterations required by the
proposed method to converge is less than 20% of what is required
by the standard conjugate gradient and neural network toolbox
algorithm.
Abstract: The paper presents a technique suitable in robot
vision applications where it is not possible to establish the object position from one view. Usually, one view pose calculation methods
are based on the correspondence of image features established at a
training step and exactly the same image features extracted at the
execution step, for a different object pose. When such a
correspondence is not feasible because of the lack of specific features
a new method is proposed. In the first step the method computes
from two views the 3D pose of feature points. Subsequently, using a
registration algorithm, the set of 3D feature points extracted at the execution phase is aligned with the set of 3D feature points extracted
at the training phase. The result is a Euclidean transform which have
to be used by robot head for reorientation at execution step.
Abstract: The prediction of Software quality during development life cycle of software project helps the development organization to make efficient use of available resource to produce the product of highest quality. “Whether a module is faulty or not" approach can be used to predict quality of a software module. There are numbers of software quality prediction models described in the literature based upon genetic algorithms, artificial neural network and other data mining algorithms. One of the promising aspects for quality prediction is based on clustering techniques. Most quality prediction models that are based on clustering techniques make use of K-means, Mixture-of-Guassians, Self-Organizing Map, Neural Gas and fuzzy K-means algorithm for prediction. In all these techniques a predefined structure is required that is number of neurons or clusters should be known before we start clustering process. But in case of Growing Neural Gas there is no need of predetermining the quantity of neurons and the topology of the structure to be used and it starts with a minimal neurons structure that is incremented during training until it reaches a maximum number user defined limits for clusters. Hence, in this work we have used Growing Neural Gas as underlying cluster algorithm that produces the initial set of labeled cluster from training data set and thereafter this set of clusters is used to predict the quality of test data set of software modules. The best testing results shows 80% accuracy in evaluating the quality of software modules. Hence, the proposed technique can be used by programmers in evaluating the quality of modules during software development.
Abstract: Cancer classification to their corresponding cohorts has been key area of research in bioinformatics aiming better prognosis of the disease. High dimensionality of gene data has been makes it a complex task and requires significance data identification technique in order to reducing the dimensionality and identification of significant information. In this paper, we have proposed a novel approach for classification of oral cancer into metastasis positive and negative patients. We have used significance analysis of microarrays (SAM) for identifying significant genes which constitutes gene signature. 3 different gene signatures were identified using SAM from 3 different combination of training datasets and their classification accuracy was calculated on corresponding testing datasets using k-Nearest Neighbour (kNN), Fuzzy C-Means Clustering (FCM), Support Vector Machine (SVM) and Backpropagation Neural Network (BPNN). A final gene signature of only 9 genes was obtained from above 3 individual gene signatures. 9 gene signature-s classification capability was compared using same classifiers on same testing datasets. Results obtained from experimentation shows that 9 gene signature classified all samples in testing dataset accurately while individual genes could not classify all accurately.
Abstract: The use of artificial neural network (ANN) modeling
for prediction and forecasting variables in water resources
engineering are being increasing rapidly. Infrastructural applications
of ANN in terms of selection of inputs, architecture of networks,
training algorithms, and selection of training parameters in different
types of neural networks used in water resources engineering have
been reported. ANN modeling conducted for water resources
engineering variables (river sediment and discharge) published in
high impact journals since 2002 to 2011 have been examined and
presented in this review. ANN is a vigorous technique to develop
immense relationship between the input and output variables, and
able to extract complex behavior between the water resources
variables such as river sediment and discharge. It can produce robust
prediction results for many of the water resources engineering
problems by appropriate learning from a set of examples. It is
important to have a good understanding of the input and output
variables from a statistical analysis of the data before network
modeling, which can facilitate to design an efficient network. An
appropriate training based ANN model is able to adopt the physical
understanding between the variables and may generate more effective
results than conventional prediction techniques.
Abstract: It is well known that Logistic Regression is the gold
standard method for predicting clinical outcome, especially
predicting risk of mortality. In this paper, the Decision Tree method
has been proposed to solve specific problems that commonly use
Logistic Regression as a solution. The Biochemistry and
Haematology Outcome Model (BHOM) dataset obtained from
Portsmouth NHS Hospital from 1 January to 31 December 2001 was
divided into four subsets. One subset of training data was used to
generate a model, and the model obtained was then applied to three
testing datasets. The performance of each model from both methods
was then compared using calibration (the χ2 test or chi-test) and
discrimination (area under ROC curve or c-index). The experiment
presented that both methods have reasonable results in the case of the
c-index. However, in some cases the calibration value (χ2) obtained
quite a high result. After conducting experiments and investigating
the advantages and disadvantages of each method, we can conclude
that Decision Trees can be seen as a worthy alternative to Logistic
Regression in the area of Data Mining.
Abstract: Researches show that probability-statistical methods application, especially at the early stage of the aviation Gas Turbine Engine (GTE) technical condition diagnosing, when the flight information has property of the fuzzy, limitation and uncertainty is unfounded. Hence the efficiency of application of new technology Soft Computing at these diagnosing stages with the using of the Fuzzy Logic and Neural Networks methods is considered. According to the purpose of this problem training with high accuracy of fuzzy multiple linear and non-linear models (fuzzy regression equations) which received on the statistical fuzzy data basis is made. For GTE technical condition more adequate model making dynamics of skewness and kurtosis coefficients- changes are analysed. Researches of skewness and kurtosis coefficients values- changes show that, distributions of GTE workand output parameters of the multiple linear and non-linear generalised models at presence of noise measured (the new recursive Least Squares Method (LSM)). The developed GTE condition monitoring system provides stage-by-stage estimation of engine technical conditions. As application of the given technique the estimation of the new operating aviation engine technical condition was made.
Abstract: One of the essential requirements in order to have a
realistic surgical simulator is real-time interaction by means of a
haptic interface is. In fact, reproducing haptic sensations increases
the realism of the simulation. However, the interaction need to be
performed in real-time, since a delay between the user action and the
system reaction reduces the user immersion. In this paper, we present
a prototype of the coronary stent implant simulator developed in the
HERMES Project; this system allows real-time interactions with a
artery by means of a specific haptic device; thus the user can
interactively navigate in a reconstructed artery and force feedback is
produced when contact occurs between the artery walls and the
medical instruments
Abstract: In today-s global and competitive market,
manufacturing companies are working hard towards improving their
production system performance. Most companies develop production
systems that can help in cost reduction. Manufacturing systems
consist of different elements including production methods,
machines, processes, control and information systems. Human issues
are an important part of manufacturing systems, yet most companies
do not pay sufficient attention to them. In this paper, a workforce
planning (WP) model is presented. A non-linear programming model
is developed in order to minimize the hiring, firing, training and
overtime costs. The purpose is to determine the number of workers
for each worker type, the number of workers trained, and the number
of overtime hours. Moreover, a decision support system (DSS) based
on the proposed model is introduced using the Excel-Lingo software
interfacing feature. This model will help to improve the interaction
between the workers, managers and the technical systems in
manufacturing.
Abstract: We propose a fast and robust hierarchical face detection system which finds and localizes face images with a cascade of classifiers. Three modules contribute to the efficiency of our detector. First, heterogeneous feature descriptors are exploited to enrich feature types and feature numbers for face representation. Second, a PSO-Adaboost algorithm is proposed to efficiently select discriminative features from a large pool of available features and reinforce them into the final ensemble classifier. Compared with the standard exhaustive Adaboost for feature selection, the new PSOAdaboost algorithm reduces the training time up to 20 times. Finally, a three-stage hierarchical classifier framework is developed for rapid background removal. In particular, candidate face regions are detected more quickly by using a large size window in the first stage. Nonlinear SVM classifiers are used instead of decision stump functions in the last stage to remove those remaining complex nonface patterns that can not be rejected in the previous two stages. Experimental results show our detector achieves superior performance on the CMU+MIT frontal face dataset.
Abstract: In this work a surgical simulator is produced which
enables a training otologist to conduct a virtual, real-time prosthetic
insertion. The simulator provides the Ear, Nose and Throat surgeon
with real-time visual and haptic responses during virtual cochlear
implantation into a 3D model of the human Scala Tympani (ST). The
parametric model is derived from measured data as published in the
literature and accounts for human morphological variance, such as
differences in cochlear shape, enabling patient-specific pre- operative
assessment. Haptic modeling techniques use real physical data and
insertion force measurements, to develop a force model which
mimics the physical behavior of an implant as it collides with the ST
walls during an insertion. Output force profiles are acquired from the
insertion studies conducted in the work, to validate the haptic model.
The simulator provides the user with real-time, quantitative insertion
force information and associated electrode position as user inserts the
virtual implant into the ST model. The information provided by this
study may also be of use to implant manufacturers for design
enhancements as well as for training specialists in optimal force
administration, using the simulator. The paper reports on the methods
for anatomical modeling and haptic algorithm development, with
focus on simulator design, development, optimization and validation.
The techniques may be transferrable to other medical applications
that involve prosthetic device insertions where user vision is
obstructed.
Abstract: Artificial Neural Network (ANN) has been
extensively used for classification of heart sounds for its
discriminative training ability and easy implementation. However, it
suffers from overparameterization if the number of nodes is not
chosen properly. In such cases, when the dataset has redundancy
within it, ANN is trained along with this redundant information that
results in poor validation. Also a larger network means more
computational expense resulting more hardware and time related
cost. Therefore, an optimum design of neural network is needed
towards real-time detection of pathological patterns, if any from heart
sound signal. The aims of this work are to (i) select a set of input
features that are effective for identification of heart sound signals and
(ii) make certain optimum selection of nodes in the hidden layer for a
more effective ANN structure. Here, we present an optimization
technique that involves Singular Value Decomposition (SVD) and
QR factorization with column pivoting (QRcp) methodology to
optimize empirically chosen over-parameterized ANN structure.
Input nodes present in ANN structure is optimized by SVD followed
by QRcp while only SVD is required to prune undesirable hidden
nodes. The result is presented for classifying 12 common
pathological cases and normal heart sound.