Methods and applications of artificial and computational intelligence
Post on 18-Dec-2016
dstructure of a video surveillance system able to detect, recognize
dangerous situations at level crossings.The authors of the paper Amethodology for the characteriza-
tion of ow conductivity through the identication of commu-nities in samples of fractured rocks propose the novel approachto characterization of possible ow conductivity in naturally frac-tured reservoir based on analysis of fracture networks obtainedfrom rock images. The proposed methodology transforms the frac-ture image into a graph and applies the graph analysis algorithms
sis, which minimizes the reconstruction error of the training
studies the problem of selecting conguration of complex multi-parametric computational Air Quality Model (AQM) which simu-lates the atmospheric transformations of gases and aerosols emit-ted into the atmosphere. A method for selecting Pareto optimalcongurations of AQM suitable for future air quality studies is pro-posed when several traditional measures and expert evaluations ofperformance of the model are used as criteria.
The paper An adaptive single-point algorithm for globalnumerical optimization describes a novel algorithm for global
Expert Systems with Applications xxx (2013) xxxxxx
.estudied in social and communication networks analysis.and evaluate potentially dangerous situations in level crossingenvironments. The proposed methodology includes different mod-els and methods of moving objects detection and separation, mul-ti-object tracking, trajectory prediction and evaluation of
documents used to compute a low-rank category transformationmatrix.
The paper Multi-criteria selection of an Air Quality Modelconguration based on quantitative and linguistic evaluationsEditorial
Methods and applications of articial an
This Special Issue is based on the expanded versions of selectedpapers presented during 11th MICAI, Mexican International Con-ference on Articial Intelligence, October 27 November 4, 2012,San Luis Potos, Mexico. MICAI conferences are organized annuallyby the Mexican Society for Articial Intelligence (SMIA). This Spe-cial Issue starts with the Invited Paper written by Francisco J. Can-tu-Ortiz, Past President of SMIA (19972000) and Conference Chairof the 1st MICAI, organized in Acapulco, Mexico in 2000. In his pa-per Advancing articial intelligence research and dissemina-tion through conference series: benchmark, scientic impactand the MICAI experience he presents an overview, analysisand benchmark of the best-known articial intelligence confer-ences, including MICAI, IJCAI, AAAI, ECAI, IBERAMIA, AAJCAI andPRICAI. Conference trends and future developments are alsoexplained.
Other papers of the Special Issue can be joined in several groupsaccording to the application areas or proposed methods: imageprocessing, medical applications, speech recognition and naturallanguage processing, global optimization with adaptive mutationstrategy, multi-criteria decision making and support vectors ma-chines, analysis of purchasing behavior of equipment manufactur-ers, workow verication and intruder detection.
During the last 2-3 years, the interest in RGB-D sensors anddepth imaging technology is dramatically increasing due to thewide range of industrial applications. The paper Evolutionaryjoint selection to improve human action recognition withRGB-D devices proposes an evolutionary algorithm to improvehuman action recognition with RGB-D cameras. The proposedreduction of the size of the feature vector used in the descriptionof human skeletons and actions by the evolutionary algorithmhas an impact on the processing rate, since the recognition is per-formed faster. This will allow the development of more complexrecognition algorithms and their application in real-time.
The paper A novel evidence based model for detecting dan-gerous situations in level crossing environments describes the
Contents lists avai
journal homepage: www0957-4174/$ - see front matter 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.eswa.2013.08.007computational intelligence
The paper Two-steps learning of fuzzy cognitive maps forprediction and knowledge discovery on the HIV-1 drug resis-tance proposes a model based on fuzzy cognitive maps for analyz-ing the behavior of the HIV-1 protease protein. A two stepslearning algorithm using swarm intelligence for optimizing themodeling parameters is introduced.
The paper Evolutionary computation in the identication ofrisk factors. Case of TRALI presents an evolutionary algorithm fordetermining factors associated with transfusion related acute lunginjury (TRALI). This algorithm is combined with the pattern recog-nition technique based on the concept of testor. A strategy for thetreatment of patients that should be transfused is proposed.
The authors of the paper Evolutionary approach for integra-tion of multiple pronunciation patterns for enhancement ofdysarthric speech recognition propose the method of enhance-ment of dysarthric speech recognition based on integration of mul-tiple pronunciation patterns. This integration is performed byweighting the responses of an Automatic Speech Recognition sys-tem for different language model restrictions. The weights are esti-mated by a genetic algorithm that also optimizes the structure ofthe implementation technique based on discrete hidden Markovmodels.
The paper Syntactic N-grams as machine learning featuresfor natural language processing describes a novel idea in naturallanguage processing consisting in construction of n-grams on syn-tactic trees. It allows to introduce the syntactic information intomachine learning methods. The results of the application of syntac-tic n-grams to the task of the automatic authorship attribution arepresented and it is shown that syntactic n-grams have better per-formance than traditional n-grams.
The authors of the paper Minimizer of the reconstruction er-ror for multi-class document categorization introduce and val-idate an approach for single-label multi-class documentcategorization based on text content features. The proposed ap-proach uses the statistical property of principal component analy-
le at ScienceDirect
lsevier .com/locate /eswa
numerical optimization, called Simple Adaptive Climbing (SAC).SAC is a simple efcient single-point approach with adaptive muta-tion process that shares many similarities with optimization heu-ristics such as Random Walk, Gradient Descent, Hill-Climbingand Differential Evolution. Test results on 15 well-known uncon-strained problems conrm that SAC is competitive against state-of-the-art approaches such as micro-Particle Swarm optimization,CMA-ES or Simple Adaptive Differential Evolution.
The paper Quadratic optimization ne tuning for the sup-port vector machines learning phase presents a comparativeanalysis of specic mathematical programming implementationtechniques of the quadratic optimization problem based on thesupport vector machines learning process. The authors proposeseveral optimization strategies of general decomposition of thequadratic optimization problem to treat large-scale applications.
The authors of the paper Characterization and analysis ofsales data for the semiconductor market: an expert system ap-proach propose the methodology for predicting purchasingbehavior in the semiconductor market. Chip purchasing policiesof the Original Equipment Manufacturers (OEMs) of laptop com-puters are characterized by similarity measures and probabilisticrules extracted from data of products bought by the OEMs in thesemiconductor market. The characterization of the purchasingbehavior of the OEMs involves multiple overlapping productslaunched into the market over time.
The authors of the paper On verication of nested workowswith extra constraints: from theory to practice formalize theproblem of workow verications as the problem of verifying thatthere exists a feasible process for each task in the workow. This
ows with extra constraints that is based on constraint satisfactiontechniques and exploits an incremental temporal reasoningalgorithm.
The paper The Windows-Users and -Intruder simulationsLogs dataset (WUIL): An experimental framework for masquer-ade detection mechanisms introduces a new masquerade data-set, called WUIL, which, unlike existing datasets, involves morefaithful masquerade attempts. WUIL exploits the hypothesis thatthe way in which a user navigates le system structure can neatlyseparate a masquerade attack. The authors argue that this ap-proach, based on le system navigation, provides a richer means,and at a higher-level of abstraction, for building novel models formasquerade detection.
The guest editors would like to thank Dr. Jay Liebowitz, theEditor-in-Chief of the Expert Systems with Applications, for thesupport in publishing this Special Issue.
Ildar BatyrshinResearch Program of Applied Mathematics and Computations,
Mexican Petroleum Institute, Eje Central Lzaro Crdenas Norte, 152,Col. San Bartolo Atepehuacan, Mexico, D.F., C.P. 07730E-mail addresses: email@example.com, firstname.lastname@example.org
Miguel Gonzalez-MendozaGraduate Programs on Computer Sciences, Tecnolgico de Monterrey,Campus Estado de Mxico, Carretera Lago de Guadalupe, Km.3.5, CP
52179 Atizapn de Zaragoza. Estado de Mxico, MexicoE-mail addresses: email@example.com, firstname.lastname@example.org
2 Editorial / Expert Systems with Applications xxx (2013) xxxxxxproblem is tractable for nested workows, e.g. for workows ofbusiness and manufacturing processes, with a hierarchical struc-ture similar to hierarchical task networks in planning. NP com-pleteness of the workow verication problem is studied. Thepaper presents a workow verication algorithm for nested work-Available online xxxx
Methods and applications of artificial and computational intelligence