-*- mode: org -*-
Miscellaneous. Diverse topics different from main ones.
Natural language processing, and other minor subfields:

*  36(<- 87): Fingerprint Recognition by Multi-objective Optimization PSO Hybrid with SVM

Researchers put efforts to discover more efficient ways to
classification problems for a period of time. Recent years, the
support vector machine (SVM) becomes a well-popular intelligence
algorithm developed for dealing this kind of problem. In this paper,
we used the core idea of multi-objective optimization to transform SVM
into a new form. This form of SVM could help to solve the situation:
in tradition, SVM is usually a single optimization equation, and
parameters for this algorithm can only be determined by user's
experience, such as penalty parameter. Therefore, our algorithm is
developed to help user prevent from suffering to use this algorithm in
the above condition. We use multi-objective Particle Swarm
Optimization algorithm in our research and successfully proved that
user do not need to use trial - and - error method to determine
penalty parameter C. Finally, we apply it to NIST-4 database to assess
our proposed algorithm feasibility, and the experiment results shows
our method can have great results as we expect.

2014





*  70(<-356): Multi-objective uniform design as a SVM model selection tool for face recognition

The primary difficulty of support vector machine (SVM) model selection
is heavy computational cost, thus it is difficult for current model
selection methods to be applied in face recognition. Model selection
via uniform design can effectively alleviate the computational cost,
but its drawback is that it adopts a single objective criterion which
can not always guarantee the generalization capacity. The sensitivity
and specificity as multi-objective criteria have been proved of better
performance and can provide a means for obtaining more realistic
models. This paper first proposes a multi-objective uniform design
(MOUD) search method as a SVM model selection tool, and then applies
this optimized SVM classifier to face recognition. Because of
replacing single objective criterion with multi-objective criteria and
adopting uniform design to seek experimental points that uniformly
scatter on whole experimental domain, MOUD can reduce the
computational cost and improve the classification ability
simultaneously. The experiments are executed on UCI benchmark, and on
Yale and CAS-PEAL-R1 face databases. The experimental results show
that the proposed method outperforms other model search methods
significantly, especially for face recognition. (C) 2010 Elsevier
Ltd. All rights reserved.

2011



*  75(<-608): Selection of optimal features for iris recognition

Iris recognition is a prospering biometric method, but some technical
difficulties still exist. This paper proposes an iris recognition
method based on selected optimal features and statistical learning. To
better represent the variation details in irises, we extract features
from both spatial and frequency domain. Multi-objective genetic
algorithm is then employed to optimize the features. Next step is
doing classification of the optimal feature sequence. SVM has recently
generated a great interest in the community of machine learning due to
its excellent generalization performance in a wide variety of learning
problems. We modified traditional SVM as non-symmetrical support
vector machine to satisfy the different security requirements in iris
recognition applications. Experimental data shows that the selected
feature sequence represents the variation details of the iris patterns
properly.

2005






*  86(<-213): Simulated annealing based classifier ensemble techniques: Application to part of speech tagging

Part-of-Speech (PoS) tagging is an important pipelined module for
almost all Natural Language Processing (NLP) application areas. in
this paper we formulate PoS tagging within the frameworks of single
and multi-objective optimization techniques. At the very first step we
propose a classifier ensemble technique for PoS tagging using the
concept of single objective optimization (SOO) that exploits the
search capability of simulated annealing (SA). Thereafter we devise a
method based on multiobjective optimization (MOO) to solve the same
problem, and for this a recently developed multiobjective simulated
annealing based technique, AMOSA, is used. The characteristic features
of AMOSA are its concepts of the amount of domination and archive in
simulated annealing, and situation specific acceptance
probabilities. We use Conditional Random Field (CRF) and Support
Vector Machine (SVM) as the underlying classification methods that
make use of a diverse set of features, mostly based on local contexts
and orthographic constructs. We evaluate our proposed approaches for
two Indian languages, namely Bengali and Hindi. Evaluation results of
the single objective version shows the overall accuracy of 88.92% for
Bengali and 87.67% for Hindi. The MOO based ensemble yields the
overall accuracies of 90.45% and 89.88% for Bengali and Hindi,
respectively. (C) 2012 Elsevier B.V. All rights reserved.

2013



*  87(<-228): Combining multiple classifiers using vote based classifier ensemble technique for named entity recognition

In this paper, we pose the classifier ensemble problem under single
and multiobjective optimization frameworks, and evaluate it for Named
Entity Recognition (NER), an important step in almost all Natural
Language Processing (NLP) application areas. We propose the solutions
to two different versions of the ensemble problem for each of the
optimization frameworks. We hypothesize that the reliability of
predictions of each classifier differs among the various output
classes. Thus, in an ensemble system it is necessary to find out
either the eligible classes for which a classifier is most suitable to
vote (i.e., binary vote based ensemble) or to quantify the amount of
voting for each class in a particular classifier (i.e., real vote
based ensemble). We use seven diverse classifiers, namely Naive Bayes,
Decision Tree (DT), Memory Based Learner (MBL), Hidden Markov Model
(HMM), Maximum Entropy (ME), Conditional Random Field (CRF) and
Support Vector Machine (SVM) to build a number of models depending
upon the various representations of the available features that are
identified and selected mostly without using any domain knowledge
and/or language specific resources. The proposed technique is
evaluated for three resource-constrained languages, namely Bengali,
Hindi and Telugu. Results using multiobjective optimization (MOO)
based technique yield the overall recall, precision and F-measure
values of 94.21%, 94.72% and 94.74%, respectively for Bengali, 99.07%,
90.63% and 94.66%, respectively for Hindi and 82.79%, 95.18% and
88.55%, respectively for Telugu. Results for all the languages show
that the proposed MOO based classifier ensemble with real voting
attains the performance level which is superior to all the individual
classifiers, three baseline ensembles and the corresponding single
objective based ensemble. (C) 2012 Elsevier B.V. All rights reserved.

2013



*  88(<-256): Combining feature selection and classifier ensemble using a multiobjective simulated annealing approach: application to named entity recognition

In this paper, we propose a two-stage multiobjective-simulated
annealing (MOSA)-based technique for named entity recognition
(NER). At first, MOSA is used for feature selection under two
statistical classifiers, viz. conditional random field (CRF) and
support vector machine (SVM). Each solution on the final Pareto
optimal front provides a different classifier. These classifiers are
then combined together by using a new classifier ensemble technique
based on MOSA. Several different versions of the objective functions
are exploited. We hypothesize that the reliability of prediction of
each classifier differs among the various output classes. Thus, in an
ensemble system, it is necessary to find out the appropriate weight of
vote for each output class in each classifier. We propose a MOSA-based
technique to determine the weights for votes automatically. The
proposed two-stage technique is evaluated for NER in Bengali, a
resource-poor language, as well as for English. Evaluation results
yield the highest recall, precision and F-measure values of 93.95,
95.15 and 94.55 %, respectively for Bengali and 89.01, 89.35 and 89.18
%, respectively for English. Experiments also suggest that the
classifier ensemble identified by the proposed MOO-based approach
optimizing the F-measure values of named entity (NE) boundary
detection outperforms all the individual classifiers and four
conventional baseline models.

2013



*  89(<-332): A multiobjective simulated annealing approach for classifier ensemble: Named entity recognition in Indian languages as case studies

In this paper, we propose a simulated annealing (SA) based
multiobjective optimization (MOO) approach for classifier
ensemble. Several different versions of the objective functions are
exploited. We hypothesize that the reliability of prediction of each
classifier differs among the various output classes. Thus, in an
ensemble system, it is necessary to find out the appropriate weight of
vote for each output class in each classifier. Diverse classification
methods such as Maximum Entropy (ME), Conditional Random Field (CRF)
and Support Vector Machine (SVM) are used to build different models
depending upon the various representations of the available
features. One most important characteristics of our system is that the
features are selected and developed mostly without using any deep
domain knowledge and/or language dependent resources. The proposed
technique is evaluated for Named Entity Recognition (NER) in three
resource-poor Indian languages, namely Bengali, Hindi and
Telugu. Evaluation results yield the recall, precision and F-measure
values of 93.95%, 95.15% and 94.55%, respectively for Bengali, 93.35%,
92.25% and 92.80%, respectively for Hindi and 84.02%, 96.56% and
89.85%, respectively for Telugu. Experiments also suggest that the
classifier ensemble identified by the proposed MOO based approach
optimizing the F-measure values of named entity (NE) boundary
detection outperforms all the individual models, two conventional
baseline models and three other MOO based ensembles. (C) 2011 Elsevier
Ltd. All rights reserved.

2011




* 145(<-187): Software Effort Estimation as a Multiobjective Learning Problem

Ensembles of learning machines are promising for software effort
estimation (SEE), but need to be tailored for this task to have their
potential exploited. A key issue when creating ensembles is to produce
diverse and accurate base models. Depending on how differently
different performance measures behave for SEE, they could be used as a
natural way of creating SEE ensembles. We propose to view SEE model
creation as a multiobjective learning problem. A multiobjective
evolutionary algorithm (MOEA) is used to better understand the
tradeoff among different performance-measures by creating SEE models
through the simultaneous optimisation of these measures. We show that
the performance measures behave very differently, presenting sometimes
even opposite trends. They are then used as a source of diversity for
creating SEE ensembles. A good tradeoff among different measures can
be obtained by using an ensemble of MOEA solutions. This ensemble
performs similarly or better than a model that does not consider these
measures explicitly. Besides, MOEA is also flexible, allowing emphasis
of a particular measure if desired. In conclusion, MOEA can be used to
better understand the relationship among performance measures and has
shown to be very effective in creating SEE models.

2013


* 153(<-626): A fuzzy multi-criteria decision approach for software development strategy selection

This study proposes a methodology to improve the quality of
decision-making in the software development project under uncertain
conditions. To deal with the uncertainty and vagueness from subjective
perception and experience of humans in the decision process, a
methodology based on the extent fuzzy analytic hierarchy process
modeling to assess the adequate economic ( tangible) and quality (
intangible) balance is applied. Two key factors of economic and
quality are evaluated separately by fuzzy approaches and both factors'
estimates are combined to obtain the preference degree associated with
each software development project strategy alternative for selecting
the most appropriate one. Using the proposed approach, the ambiguities
involved in the assessment data can be effectively represented and
processed to assure a more convincing and effective
decision-making. Finally, a real case-study is given to demonstrate
the potential of the methodology.

2004




* 275(<- 89): A group decision classifier with particle swarm optimization and decision tree for analyzing achievements in mathematics and science

Group decision making is a multi-criteria decision-making method
applied in many fields. However, the use of group decision-making
techniques in multi-class classification problems and rule generation
is not explored widely. This investigation developed a group decision
classifier with particle swarm optimization (PSO) and decision tree
(GDCPSODT) for analyzing students' mathematic and scientific
achievements, which is a multi-class classification problem involving
rule generation. The PSO technique is employed to determine weights of
condition attributes; the decision tree is used to generate rules. To
demonstrate the performance of the developed GDCPSODT model, other
classifiers such as the Bayesian classifier, the k-nearest neighbor
(KNN) classifier, the back propagation neural networks classifier with
particle swarm optimization (BPNNPSO) and the radial basis function
neural networks classifier with PSO (RBFNNPSO) are used to cope with
the same data. Experimental results indicated the testing accuracy of
GDCPSODT is higher than the other four classifiers. Furthermore, rules
and some improvement directions of academic achievements are provided
by the GDCPSODT model. Therefore, the GDCPSODT model is a feasible and
promising alternative for analyzing student-related mathematic and
scientific achievement data.

2014




* 279(<-396): Towards the selection of best neural network system for intrusion detection

Currently, network security is a critical issue because a single
attack can inflict catastrophic damages to computers and network
systems. Various intrusion detection approaches are available to
adhere to this severe issue, but the dilemma is, which one is more
suitable. Being motivated by this situation, in this paper, we
evaluate and compare different neural networks (NNs). The current work
presents an evaluation of different neural networks such as
Self-organizing map (SOM), Adaptive Resonance Theory (ART), Online
Backpropagation (OBPROP), Resilient Backpropagation (RPROP) and
Support Vector Machine (SVM) towards intrusion detection mechanisms
using Multi-criteria Decision Making (MCDM) technique. The results
indicate that in terms of performance, supervised NNs are better,
while unsupervised NNs are better regarding training overhead and
aptitude towards handling varied and coordinated
intrusion. Consequently, the combined, that is, hybrid approach of NNs
is the optimal solution in the area of intrusion detection. The
outcome of this work may help and guide the security implementers in
two possible ways, either by using the results directly obtained in
this paper or by extracting the results using other similar mechanism,
but on different intrusion detection systems or approaches.

2010


* 400(<-581): Prediction of human behaviour using artificial neural networks

This paper contributes to the analysis and prediction of deviate
intentional behaviour of human operators in Human-Machine Systems
using Artificial Neural Networks that take uncertainty into
account. Such deviate intentional behaviour is a particular violation,
called Barrier Removal. The objective of the paper is to propose a
predictive Benefit-Cost-Deficit model that allows a multi-reference,
multi-factor and multi-criterion evaluation. Human operator
evaluations can be uncertain. The uncertainty of their subjective
judgements is therefore integrated into the prediction of the Barrier
Removal. The proposed approach is validated on a railway application,
and the prediction convergence of the uncertainty-integrating model is
demonstrated.

2006




* 419(<-109): INTELLIGENT ADAPTIVE MULTI-PARAMETER MIGRATION MODEL FOR LOAD BALANCING VIRTUALIZED CLUSTER OF SERVERS

The most important benefit of virtualization is to get a load balanced
environment through Virtual Machine (VM) migration. Performance of
clustered services such as Average Response Time is reduced through
intelligent VM migration decision. Migration depends on a variety of
criteria like resource usage (CPU usage, RAM usage, Network Usage,
etc.) and demand of machines (Physical (PM) and Virtual (VM)). This is
a multi-criteria migration problem that evaluates, compares and sorts
a set of PMs and VMs on the basis of parameters affected on migration
process. But, which parameter(s) has dominant role over cluster
performance in each time window? How can we determine weight of
parameters over oncoming time slots? Current migration algorithms do
not consider time-dependent variable weights of parameters. These
studies assume fixed weight for each parameter over a wide range of
time intervals. This approach leads to imprecise prediction of
recourse demand of each server. Our paper presents a new Intelligent
and Adaptive Multi Parameter migration-based resource manager (IAMP)
for virtualized data centres and clusters with a novel Artificial
Neural Network (ANN)-based weighting analysis named Error Number of
Parameter Omission (ENPO). In each time slot, weight of parameters is
recalculated and non-important ones will be attenuated in ranking
process. We characterized the parameters affecting cluster performance
and used hot migration with emphasis on cluster of servers in XEN
virtualization platform. The experimental results based on workloads
composed of real applications, indicate that IAMP management framework
is feasible to improve the performance of the virtualized cluster
system up to 23% compared to current algorithms. Moreover, it reacts
more quickly and eliminates hot spots because of its full dynamic
monitoring algorithm.

2014



* 444(<-259): Concurrent Optimization of Computationally Learned Stylistic Form and Functional Goals

Great design often results from intelligently balancing tradeoffs and
leveraging of synergies between multiple product goals. While the
engineering design community has numerous tools for managing the
interface between functional goals in products, there are currently no
formalized methods to concurrently optimize stylistic form and
functional requirements. This research develops a method to coordinate
seemingly disparate but highly related goals of stylistic form and
functional constraints in computational design. An artificial neural
network (ANN) based machine learning system was developed to model
surveyed consumer judgments of stylistic form quantitatively. Coupling
this quantitative model of stylistic form with a genetic algorithm
(GA) enables computers to concurrently account for multiple objectives
in the domains of stylistic form and more traditional functional
performance evaluation within the same quantitative framework. This
coupling then opens the door for computers to automatically generate
products that not only work well but also convey desired styles to
consumers. [DOI: 10.1115/1.4007304]

2012




* 453(<-627): Artificial neural network for violation analysis

Barrier removal (BR) is a safety-related violation, and it can be
analyzed in terms of benefits, costs, and potential deficits. In order
to allow designers to integrate BR into the risk analysis during the
initial design phase or during re-design work, we propose a
connectionist method integrating self-organizing maps (SOM). The basic
SOM is an artificial neural network that, on the basis of the
information contained in a multi-dimensional space, generates a space
of lesser dimensions. Three algorithms-Unsupervised SOM, Supervised
SOM, and Hierarchical SOM-have been developed to permit BR
classification and prediction in terms of the different criteria. The
proposed method can be used, on the one hand, to foresee/predict the
possibility level of a new/changed barrier (prospective analysis), and
on the other hand, to synthetically regroup/rearrange the BR of a
given human-machine system (retrospective analysis). We applied this
method to the BR analysis of an experimental railway simulator, and
our preliminary results are presented here. (C) 2003 Elsevier Ltd. All
rights reserved.

2004




* 468(<-653): Controlling the gray component with Pareto-optimal color space transformations

The Pareto-optimal approach to color management, presented previously
by the authors, is further developed to allow for direct conversion
into CMYK with complete user control over gray component replacement
(GCR). The Pareto-optimal formulation unifies a number of strategies
for transforming image data into CMYK: specification of arbitrary GCR,
conversion using only chromatic inks, and conversion using at most two
chromatic inks and black ink. Use of the black printer is analyzed in
terms of extending the CMY gamut and replacing chromatic inks. The
program NeuralColor is used to implement the Pareto-optimal
formulation, providing data for in-depth analysis of the various
conversion methods. NeuralColor accurately models the transformation
from CIELAB to CMYK using artificial neural networks. Prints obtained
using NeuralColor are accurate within 2 to 4 DeltaE*(ab) across all
levels of GCR.

2002


* 469(<-657): Pareto-optimal formulations for cost versus colorimetric accuracy trade-offs in printer color management

Color management for the printing of digital images is a challenging task, due primarily to nonlinear ink- mixing behavior and the presence of redundant solutions for print devices with more than three inks. Algorithms for the conversion of image data to printer- specific format are typically designed to achieve a single predetermined rendering intent, such as colorimetric accuracy. In the present paper we present two CIELAB to CMYK color conversion schemes based on a general Pareto-optimal formulation for printer color management. The schemes operate using a 149- color characterization data set selected to efficiently capture the entire CMYK gamut. The first scheme uses artificial neural networks as transfer functions between the CIELAB and CMYK spaces. The second scheme is based on a reformulation of tetrahedral interpolation as an optimization problem. Characterization data are divided into tetrahedra for the interpolation- based approach using the program Qhull, which removes the common restriction that characterization data be well organized. Both schemes offer user control over trade- off problems such as cost versus reproduction accuracy, allowing for user- specified print objectives and the use of constraints such as maximum allowable ink and maximum allowable DeltaE*(ab). A formulation for minimization of ink is shown to be particularly favorable, integrating both clipping and gamut compression features into a single methodology. Codes developed as applications of these schemes were used to convert several CIELAB Tiff images to CMYK format, providing both qualitative and quantitative verification of the Pareto-optimal approach. Prints of the MacBeth ColorChecker (tm) chart were accurate within approximately to 3 DeltaE*(ab) for in- gamut colors. Modifications to this approach are presented that offer user control over grey component replacement and provide additional options for rendering intent.

2002


* 500(<- 23): Correction of Gravimetric Geoid Using Symbolic Regression

In this study, the problem of geoid correction based on GPS
ellipsoidal height measurements is solved via symbolic regression
(SR). In this case, when the quality of the approximation is
overriding, SR employing Keijzer expansion to generate initial trial
function population can supersede traditional techniques, such as
parametric models and soft computing models (e.g., artificial neural
network approach with different activation functions). To demonstrate
these features, numerical computations for correction of the Hungarian
geoid have been carried out using the DataModeler package of
Mathematica. Although the proposed SR method could reduce the average
error to a level of 1-2 cm, it has two handicaps. The first one is the
required high computation power, which can be eased by the employment
of parallel computation via multicore processor. The second one is the
proper selection of the initial population of the trial
functions. This problem may be solved via intelligent generation
technique of this population (e.g., Keijzer-expansion).

2015







* 503(<-278): Functionality defense through diversity: a design framework to multitier systems

Diversification is one of the most effective approaches to defend
multitier systems against attacks, failure, and accidents. However,
designing such a system with effective diversification is a
challenging task because of stochastic user and attacker behaviors,
combinatorial-explosive solution space, and multiple conflicting
design objectives. In this study, we present a systematic framework
for exploring the solution space, and consequently help the designer
select a satisfactory system solution. A simulation model is employed
to evaluate design solutions, and an artificial neural network is
trained to approximate the behavior of the system based on simulation
output. Guided by a trained neural network, a multi-objective
evolutionary algorithm (MOEA) is proposed to search the solution space
and identify potentially good solutions. Our MOEA incorporates the
concept of Herbert Simon's satisficing. It uses the decision maker's
aspiration levels for system performance metrics as its search
direction to identity potentially good solutions. Such solutions are
then evaluated via simulation. The newly-obtained simulation results
are used to refine the neural network. The exploration process stops
when the result converges or a satisfactory solution is found. We
demonstrate and validate our framework using a design case of a
three-tier web system.

2012