Foundations of Computational Intelligence Volume 5: Function Approximation and Classification

378 Pages·2009·4.53 MB·English

by  Guo-Zheng Li

AjithAbraham,Aboul-EllaHassanien,andVa´clavSna´ˇsel(Eds.)
FoundationsofComputationalIntelligenceVolume5
StudiesinComputationalIntelligence,Volume205
Editor-in-Chief
Prof.JanuszKacprzyk
SystemsResearchInstitute
PolishAcademyofSciences
ul.Newelska6
01-447Warsaw
Poland
E-mail:[email protected]
Furthervolumesofthisseriescanbefoundonour Vol.195.VivekBannoreandLeszekSwierkowski
homepage:springer.com Iterative-InterpolationSuper-ResolutionImage
Reconstruction:
Vol.181.GeorgiosMiaoulisandDimitriPlemenos(Eds.) AComputationallyEfficientTechnique,2009
IntelligentSceneModellingInformationSystems,2009 ISBN978-3-642-00384-4
ISBN978-3-540-92901-7
Vol.196.ValentinaEmiliaBalas,Ja´nosFodorand
Vol.185.AnthonyBrabazonandMichaelO’Neill(Eds.) AnnamáriaR.Va´rkonyi-Ko´czy(Eds.)
NaturalComputinginComputationalFinance,2009 SoftComputingBasedModeling
ISBN978-3-540-95973-1 inIntelligentSystems,2009
ISBN978-3-642-00447-6
Vol.186.Chi-KeongGohandKayChenTan
Vol.197.MauroBirattari
EvolutionaryMulti-objectiveOptimizationinUncertain
TuningMetaheuristics,2009
Environments,2009
ISBN978-3-642-00482-7
ISBN978-3-540-95975-5
Vol.198.Efre´nMezura-Montes(Ed.)
Vol.187.MitsuoGen,DavidGreen,OsamuKatai,BobMcKay,
Constraint-HandlinginEvolutionaryOptimization,2009
AkiraNamatame,RuhulA.SarkerandByoung-TakZhang
ISBN978-3-642-00618-0
(Eds.)
IntelligentandEvolutionarySystems,2009
Vol.199.KazumiNakamatsu,GloriaPhillips-Wren,
ISBN978-3-540-95977-9
LakhmiC.Jain,andRobertJ.Howlett(Eds.)
Vol.188.AgustínGutie´rrezandSantiagoMarco(Eds.) NewAdvancesinIntelligentDecisionTechnologies,2009
BiologicallyInspiredSignalProcessingforChemicalSensing, ISBN978-3-642-00908-2
2009
ISBN978-3-642-00175-8 Vol.200.DimitriPlemenosandGeorgiosMiaoulisVisual
ComplexityandIntelligentComputerGraphicsTechniques
Vol.189.SallyMcClean,PeterMillard,EliaEl-Darziand Enhancements,2009
ChrisNugent(Eds.) ISBN978-3-642-01258-7
IntelligentPatientManagement,2009
ISBN978-3-642-00178-9 Vol.201.Aboul-EllaHassanien,AjithAbraham,
AthanasiosV.Vasilakos,andWitoldPedrycz(Eds.)
Vol.190.K.R.Venugopal,K.G.SrinivasaandL.M.Patnaik FoundationsofComputationalIntelligenceVolume1,2009
SoftComputingforDataMiningApplications,2009 ISBN978-3-642-01081-1
ISBN978-3-642-00192-5
Vol.202.Aboul-EllaHassanien,AjithAbraham,
Vol.191.ZongWooGeem(Ed.) andFranciscoHerrera(Eds.)
Music-InspiredHarmonySearchAlgorithm,2009 FoundationsofComputationalIntelligenceVolume2,2009
ISBN978-3-642-00184-0 ISBN978-3-642-01532-8
Vol.192.AgusBudiyono,BambangRiyantoandEndra Vol.203.AjithAbraham,Aboul-EllaHassanien,
Joelianto(Eds.) PatrickSiarry,andAndriesEngelbrecht(Eds.)
IntelligentUnmannedSystems:TheoryandApplications,2009 FoundationsofComputationalIntelligenceVolume3,2009
ISBN978-3-642-00263-2 ISBN978-3-642-01084-2
Vol.193.RaymondChiong(Ed.) Vol.204.AjithAbraham,Aboul-EllaHassanien,and
Nature-InspiredAlgorithmsforOptimisation,2009 Andre´PoncedeLeonF.deCarvalho(Eds.)
ISBN978-3-642-00266-3 FoundationsofComputationalIntelligenceVolume4,2009
ISBN978-3-642-01087-3
Vol.194.IanDempsey,MichaelO’NeillandAnthony
Brabazon(Eds.) Vol.205.AjithAbraham,Aboul-EllaHassanien,and
FoundationsinGrammaticalEvolutionforDynamic VáclavSnášel(Eds.)
Environments,2009 FoundationsofComputationalIntelligenceVolume5,2009
ISBN978-3-642-00313-4 ISBN978-3-642-01535-9
AjithAbraham,Aboul-Ella Hassanien,
andVa´clav Sna´ˇsel (Eds.)
Foundations of Computational
Intelligence Volume 5
Function Approximation and Classification
123
Prof.AjithAbraham Prof.VaclavSnášel
MachineIntelligenceResearchLabs TechnicalUniversityOstrava
(MIRLabs) Dept.ComputerScience
ScientificNetworkforInnovationand Tr.17.Listopadu15
ResearchExcellence 70833Ostrava
P.O.Box2259 CzechRepublic
Auburn,Washington98071-2259 E-mail:[email protected]
USA
E-mail:[email protected]
Prof.Aboul-EllaHassanien
CairoUniversity
FacultyofComputersandInformation
InformationTechnologyDepartment
5AhmedZewalSt.
Orman,Giza
E-mail:[email protected]
http://www.fci.cu.edu.eg/abo/
ISBN 978-3-642-01535-9 e-ISBN978-3-642-01536-6
DOI 10.1007/978-3-642-01536-6
Studiesin Computational Intelligence ISSN1860949X
Library of Congress Control Number:Applied for
(cid:2)c 2009 Springer-VerlagBerlin Heidelberg
Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpart
of the material is concerned, specifically therights of translation, reprinting,reuse
ofillustrations, recitation,broadcasting, reproductiononmicrofilmorinanyother
way, and storage in data banks. Duplication of this publication or parts thereof is
permittedonlyundertheprovisionsoftheGermanCopyrightLawofSeptember9,
1965, in its current version, and permission for use must always be obtained from
Springer. Violations are liable to prosecution undertheGerman Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this
publication does not imply, even in the absence of a specific statement, that such
names are exempt from the relevant protective laws and regulations and therefore
free for general use.
Typeset&CoverDesign:ScientificPublishing ServicesPvt. Ltd., Chennai, India.
Printed in acid-free paper
9 8 7 6 5 4 3 2 1
springer.com
Preface
Foundations of Computational Intelligence
Volume 5: Function Approximation and Classification
Approximation theory is that area of analysis which is concerned with the ability
to approximate functions by simpler and more easily calculated functions. It is an
area which, like many other fields of analysis, has its primary roots in the mathe-
matics.The need for function approximation and classification arises in many
branches of applied mathematics, computer science and data mining in particular.
This edited volume comprises of 14 chapters, including several overview Chap-
ters, which provides an up-to-date and state-of-the art research covering the theory
and algorithms of function approximation and classification. Besides research arti-
cles and expository papers on theory and algorithms of function approximation
and classification, papers on numerical experiments and real world applications
were also encouraged.
The Volume is divided into 2 parts:
Part-I: Function Approximation and Classification – Theoretical Foundations
Part-II: Function Approximation and Classification – Success Stories and Real
World Applications
Part I on Function Approximation and Classification – Theoretical Foundations
contains six chapters that describe several approaches Feature Selection, the use
Decomposition of Correlation Integral, Some Issues on Extensions of Information
and Dynamic Information System and a Probabilistic Approach to the Evaluation
and Combination of Preferences
Chapter 1 “Feature Selection for Partial Least Square Based Dimension Reduc-
tion” by Li and Zeng investigate a systematic feature reduction framework by
combing dimension reduction with feature selection. To evaluate the proposed
framework authors used four typical data sets. Experimental results illustrate that
the proposed method improves the performance on the gene expression microarray
data in terms of accuracy.
In Chapter 2, “Classification by the Use of Decomposition of Correlation Inte-
gral” Jirina and Jirina Jr. illustrate that the correlation integral can be decomposed
into functions each related to a particular point of data space. For these functions,
VI Preface
one can use similar polynomial approximations such as the correlation integral.
The essential difference is that the value of the exponent, which would correspond
to the correlation dimension, differs in accordance to the position of the point in
question.
Chapter 3, “Investigating Neighborhood Graphs for Inducing Density Based
Clusters” by Kasahara and Nicoletti investigate the impact of seven different
ways of defining a neighborhood region between two points, when identifying
clusters in a neighborhood graph, particularly focusing on density based clusters.
On the one hand results show that the neighborhood definitions that do not employ
parameters are not suitable for inducing density based clusters. On the other hand
authors also illustrate that although essential for successfully separating density
based clusters, the parameters employed by some of the definitions need to have
their values tuned by the user.
In Chapter 4, “Some Issues on Extensions of Information and Dynamic Infor-
mation Systems” Pancerz discusses some issues on extensions of information and
dynamic information systems. Consistent and partially consistent extensions of
information and dynamic information systems are helpful in prediction problems.
On the basis of those extensions author determine possibility distributions of states
and transitions between states over a universe of discourse related to a given sys-
tem of processes.
Chapter 5, “A Probabilistic Approach to the Evaluation and Combination of
Preferences” Parracho and Anna propose a model for the process of evaluating
and combining preferences. After determining the preferences according to each
criterion, these partial preferences are combined into global ones. Different com-
bination rules are set, in a probabilistic framework. Attention is centered to the
case of indicators measured in different levels of aggregation.
In Chapter 6, “Use of the q-Gaussian Function in Radial Basis Function Net-
works” Tinos and Murta Jr. deploy q-Gaussian function as a radial basis function
in RBF Networks for pattern recognition problems. The use of q-Gaussian RBFs
allows to modify the shape of the RBF by changing the real parameter q, and to
employ radial units with different RBF shapes in a same RBF Network.
Part II on Function Approximation and Classification – Success Stories and Real
World Applications contains six chapters that describe several success stories and
real world applications onFunction Approximation and Classification
Chapter 7, “Novel biomarkers for prostate cancer revealed by(α,β)) - k-feature
sets” by Ravetti et al. present a method based on the (α,β)- k- feature set prob-
lem for identifying relevant attributes in high-dimensional datasets for classifica-
tion purposes. Using the gene expression of thousands of genes, authors illustrate
that the method can give a reduced set that can identify samples as belonging to
prostate cancer tumors or not.
Most classification problems associate a single class to each example or in-
stance. However, there are many classification tasks where each instance can be
associated with one or more classes. This group of problems represents an area
Preface VII
known as multi-label classification. In Chapter 8, “A Tutorial on Multi-Label
Classification Techniques” Carvalho and Freitas present the most frequently used
techniques to deal with these problems in a pedagogical manner, with examples
illustrating the main techniques and proposing a taxonomy of multi-label tech-
niques that highlights the similarities and differences between these techniques.
In Chapter 9, “Computational Intelligence in Biomedical Image Processing”
Bollenbeck and Seiffert show that the segmentation of biological images, charac-
terized by non-uniform image features, significantly benefits from combining
global physical models and local feature-based supervised classification. Authors
used an entropy-based voting of optimal feed-forward networks by cross-
validation architecture selection and global registration-based segmentation.
Chapter 10, “A Comparative Study of Three Graph Edit Distance Algorithms”
by Gao et al. propose two cost function-free GED algorithms. In the edge direc-
tion histogram (EDH)-based method, edit operations are involved in graph struc-
ture difference characterized by EDH and GED is converted into earth mover's
distance (EMD) of EDHs, while edit operations are involved in node distribution
difference characterized by HMM and GED is converted into KLD of HMMs in
the HMM-based method. With respect to two cost function free algorithms,
HMM-based method excels EDH-based method in classification and clustering
rate, and efficiency.
In Chapter 11, “Classification of Complex Molecules” by Torrens and Castel-
lano introduce algorithms for classification and taxonomy based on criteria, e.g.,
information entropy and its production. In molecular classification, the feasibility
of replacing a given molecule by similar ones in the composition of a complex
drug is studied.
In Chapter 12 “Intelligent finite element method and application to simulation
of behavior of soils under cyclic loading” Javad et al. present a neural network-
based finite element method for modeling of the behavior of soils under cyclic
loading. The methodology is based on the integration of a neural network in a fi-
nite element framework. In this method, a neural network is trained using experi-
mental data representing the mechanical response of material to applied load. The
trained network is then incorporated in the finite element analysis to predict the
constitutive relationships for the material.
Chapter 13, “An Empirical Evaluation of the Effectiveness of Different Types
of Predictor Attributes in Protein Function Prediction” Otero et al. present an em-
pirical evaluation of different protein representations for protein function predic-
tion in terms of maximizing predictive accuracy, investigating which type of rep-
resentation is more suitable for different levels of hierarchy.
In the last Chapter, “Genetic Selection Algorithm and Cloning for Data Mining
with GMDH Method” Jirina and Jirina Jr. generalize the idea of the genetically
modified GMDH neural network for processing multivariate data appearing in
data mining problems and to extend this type of network by cloning. Clones are
close, but not identical copies of original individuals. The new genetically modi-
fied GMDH method with cloning (GMC GMDH) has no tough layered structure.
We are very much grateful to the authors of this volume and to the reviewers
for their great efforts by reviewing and providing interesting feedback to authors
VIII Preface
of the chapter. The editors would like to thank Dr. Thomas Ditzinger (Springer
Engineering Inhouse Editor, Studies in Computational Intelligence Series),
Professor Janusz Kacprzyk (Editor-in-Chief, Springer Studies in Computational
Intelligence Series) and Ms. Heather King (Editorial Assistant, Springer Verlag,
Heidelberg) for the editorial assistance and excellent cooperative collaboration to
produce this important scientific work. We hope that the reader will share our joy
and will find it useful!
December 2008 Ajith Abraham, Trondheim, Norway
Aboul Ella Hassanien, Cairo, Egypt
Václav Snášel, Ostrava, Czech Republic
Contents
Part I: Function Approximation and Classification: Theoretical
Foundations
Feature Selectionfor Partial Least Square Based Dimension
Reduction
Guo-Zheng Li, Xue-Qiang Zeng.................................. 3
Classification by the Use of Decomposition of Correlation
Integral
Marcel Jiˇrina, Marcel Jiˇrina Jr. ................................. 39
Investigating Neighborhood Graphs for Inducing Density
Based Clusters
Viviani Akemi Kasahara, Maria do Carmo Nicoletti................ 57
Some Issues on Extensions of Information and Dynamic
Information Systems
Krzysztof Pancerz.............................................. 79
A Probabilistic Approach to the Evaluation and
Combination of Preferences
Annibal Parracho Sant’Anna .................................... 107
Use of the q-Gaussian Function in Radial Basis Function
Networks
Renato Tin´os, Luiz Ota´vio Murta Ju´nior.......................... 127