N. Bassiliades1, I. Maglogiannis2, and E. Stamatatos2
1Dept. of Informatics, Aristotle University of Thessaloniki
54124 – Thessaloniki, Greece
2Dept. of Information and Communication Systems Eng., University of the Aegean
83200 – Karlovassi, Greece

 

TABLE OF CONTENTS

1. INTRODUCTION

2. AI APPLICATIONS

3. MATHEMATICAL FOUNDATIONS

4. CLOSURE AND REMARKS

 

 

1. INTRODUCTION


Artificial Intelligence tools and methodologies have successfully been used in a wide range of fields including medical diagnosis, stock trading, robot control, law, scientific discovery, toys and many more. Frequently, when a technique reaches mainstream use it is no longer considered Artificial Intelligence, which sometimes is described as the AI effect.

One of the reasons that AI tools and applications were successfully adopted throughout the technology industry is that these tools are truly scientific, in the sense that their results are both measurable and verifiable. This success was partly due to the commitment by researchers to solid mathematical methods and rigorous scientific standards.

In this chapter we summarize some of the AI applications and mathematical foundations of AI that were developed by Greek researchers and research groups, working in Greek universities and research institutions, or working abroad. Notice that this chapter cannot be considered as an exhaustive list of Greek researchers or groups developing AI applications and foundations, but only as an indicative list, that resulted from an open call for contributions published in Hellenic AI Society's members' mailing list.

In the following sections we first review some AI applications in the areas of Artificial Life, Business, Commerce, Finance, Government, Health, Education, Engineering, Industry, Telecommunications, Web, and Pervasive Systems. Then, we include some theoretical works on the Mathematical foundations of AI and in the final section we conclude this chapter.

 

2. AI APPLICATIONS

 

2.1 ARTIFICIAL LIFE

Dr. Yannakakis at the IT-University of Copenhagen, Denmark (http://www.itu.dk/~yannakakis/), has researched the combination of the fields of artificial life and multi-agent systems and, more specifically, has studied the emergence of cooperative multi-agent spatial coordination. By observing the global performance of a group of homogeneous agents—supported by a non-global knowledge of their environment—the attempt is to extract information about the minimum size of the agent neurocontroller and the type of learning mechanism that collectively generate high performing and robust behaviors with minimal computational effort. Consequently, methodologies for obtaining controllers of minimal size have been introduced and comparative studies between supervised and unsupervised learning mechanisms for the generation of successful collective behaviors have been presented (Yannakakis, et al., 2007). Furthermore, the FlatLand prototype simulated world has been developed to serve as a machine learning benchmark (Yannakakis, et al., 2007). This case study is primarily a computer games inspired world but its main features are also biologically plausible. FlatLand demonstrated that cooperative behavior among agents, which is supported only by limited communication, appears to be necessary for the problem’s efficient solution and that learning by rewarding the behavior of agent groups constitutes a more efficient and computationally preferred generic approach than supervised learning approaches in such complex multi-agent worlds.

The Intelligence, Modelling & Computation (IMC) research group at CITY College (http://www.city.academic.gr/special/research/imc/index.html) is engaged with research in mathematical foundations and formal aspects of modelling agents and multi-agent systems and their application to novel areas such as modelling and computation of biological systems (Kefalas, et al., 2005). More specifically, the group is researching the following topics of Computational Systems’ Biology: Modelling & Computation (Complex Systems, Emergence, Neural Networks, Genetic Algorithms), Biology-inspired systems & applications, Membrane computing (P Systems). The lack of the methods used to meet the requirement for formal modelling of self-organisation, self-assembly and emergence, led the group to the development of a flexible framework, called OPERAS (Stamatopoulou, et al., 2008), which could integrate various methods. The most prominent one was inspired by Membrane Computing, a new paradigm for computation. In particular, P Systems provided the means to allow modeling of change in the organization in a multi-agent system OPERAS (Stamatopoulou, et al., 2007). This is a characteristic of all biology and biologically–inspired systems that include colonies of social insects, flocking, schooling, and, in general, swarm intelligence. The IMC group is in close collaboration with the Computational Biology research groups of the Department of Computer Science of the University of Sheffield, through the South-East European Research Center (SEERC), established as a joint venture in Thessaloniki between CITY College and the University of Sheffield.

 

2.2 BUSINESS-COMMERCE-FINANCE-GOVERNMENT

The Management and Decision Engineering Laboratory (MDE) (http://decision.fme.aegean.gr) of the Department of Financial and Management Engineering of the University of the Aegean (Chios), focuses on intelligent computational decision making approaches, applied in real world problems related to management, finance, strategic planning, market analysis, etc. MDE Lab has designed semi-parametric financial forecasting (NN-GARCH) models, that combine intelligent learning techniques based on neural networks, and statistical - econometric GARCH models of volatility (Thomaidis, et al., 2006a; Thomaidis, 2007). The methodology proposed can accommodate most of the stylized facts reported about financial prices or rates of return such as non linear corrections, asymmetric GARCH effects and non-gaussian errors. By jointly modeling the conditional mean and volatility of the data-generating process, the scope of NNs is extended from function approximation to density forecasting tasks and the construction of neural network models under special statistical features existing in financial and economic data is also reconsidered. Furthermore, the MDE Lab has proposed a novel methodology for intelligent statistical arbitrage based financial trading (Thomaidis, et al., 2006b), a fuzzy multicriteria decision making approach for project evaluation (Thomaidis, et al., 2006c), a nature inspired intelligence-based optimization technique for the selection of portfolio financial assets (Thomaidis, et al., 2008), and computational intelligence models for discovering behavioral patterns in securities prices (Thomaidis & Dounias, 2006; Thomaidis, 2006).

The company "DIRECTING intelligence in business" (http://www.directing.gr) focuses on study, planning and implementing of their own products-solutions (DATACTIF, ADVALID, etc...) and on providing supportive services and systems such as data migration, systems integration, data management, data warehouse and data mining. Especially in the area of data mining, DIRECTING has developed the DATACTIF® data mining suite that retrieves all enterprise data producing knowledge and prediction. Due to its user friendly interface, the simplicity of usage and the human oriented design, the suite can be used by the CEO of the company, the BI specialist, or the salesman with huge market experience but no particular knowledge of statistics and computer science. DATACTIF® is based on neural networks, fuzzy systems and genetic algorithms and provides to decision makers functions such as clustering, classification, association rules, prediction using all company data (both quantitative and qualitative). It runs on several OSs as well as DBMSs and it is connected on-line with the company's MIS (CRM, Data Warehouse, etc.), retrieving and analysing data in real time. Among the advantages of DATACTIF® is that it can be adapted in the needs and specifications of each company, it adopts the most advanced techniques in the sector of intelligent systems (neural networks, fuzzy logic and genetic algorithms), it exports immediately the knowledge from data as logical expressions (rules and correlations) and as optical representations allowing direct, simple reading from decision makers, it solves complex problems with a high level of confidence even with set of data with insufficient and/or erroneous values, it handles qualitative and quantitative elements simultaneously, it analyzes nominal and continuous variables, it's usage does not require any statistical or computing knowledge.

Dr. Tsadiras at Department of Information Technology, Alexander Technological Educational Institute of Thessaloniki, has worked on computer based decision making, especially in cases of fuzziness and/or uncertainty. To this end he has proposed Certainty Neuron Fuzzy Cognitive Maps (CNFCM) (Tsadiras & Margaritis, 1997), which have memory and decay mechanisms, features that make them suitable for highly uncertain situations. This decision making technique was studied for financial decisions, such as decisions and strategic planning in a car industry (Tsadiras & Margaritis, 1995; Tsadiras, 2005). In those applications, various “what-if” scenarios are imposed to the CNFCM system and the system predicts the corresponding consequences of these hypotheses. If the predicted consequences are positive for the car industry then the imposed scenario is supported by the system and the car industry can proceed to its application. If the consequences are negative, the scenarios are not supported and suitable strategic movements must be made in order to change the negative current situation for the car industry. Another application of CNFCMs was in the area of management decisions (Tsadiras, 2007), where e.g. the consequence of the increase/decrease of the interdependence or the size of companies’ teams to the overall quality of the work, are examined. In the area of political decisions, CNFCMs were applied to examine the consequences of specific political decisions of Turkey and other involving countries towards Turkey’s integration into the European Union (Tsadiras & Kouskouvelis, 2005). Finally, the methodology was also applied successfully in topics concerning ecology, education and accounting.

 

2.3 E-HEALTH

The Computing Systems Laboratory (Prof. G. Papakonstantinou) in the Department of Electrical and Computer Engineering NTUA has done research in this field. More specifically in the frame of a funded project (GSRT-PENED 87ΕΔ265), an expert system cell (NTUA-expert) having full theorem proving capabilities and incorporating different models of inexact reasoning, was developed. This tool was used for the development of an expert system for the interpretation of Electrocardiograms (ECGs) (Bourlas, et al., 1996). In the frame of a European project (TIDE), a sensor-aided intelligent wheel chair navigation system was developed (Katevas, et al., 1997). This system was based on Artificial Neural Nets learning techniques. The 24-hour blood pressure as well as the heart-rate variability was analyzed using Kohonen's self-organizing neural networks, in order to predict future heart failure (Tambouratzis, et al., 2002).

A large part of applied AI research in the Department of Computer Science, University of Ioannina (contact Person: Aristidis Likas), is focused on the field of Bioinformatics. More specifically, the research group in UoI has developed the Greedy-MEME algorithm, a statistical incremental learning algorithm for the discovery of motifs in biological (either DNA or protein) sequences (Blekas, et al., 2003). The discovered motifs can subsequently be used as features for the classification of protein families (Blekas, et al., 2005a). They have also developed a novel method for the analysis of DNA microarray images that uses an efficient technique for microarray gridding to locate the spots in the microarray images and then employs mixture modeling methods to analyse each individual spot image (Blekas, et al., 2005b).

Artificial Intelligence Group (AI GROUP) (contact Person: Ioannis Hatzilygeroudis ) in the Laboratory of Graphics, Multimedia & GIS - Department of Computer Engineering & Informatics - University of Patras (http://mmlab.ceid.upatras.gr/aigroup/) have also used hybrid intelligent approaches for medical diagnosis. A first approach, that combines a rule-based approach with a genetic algorithm, concerns a hybrid intelligent system (called GADIS) for diagnosis of male impotence. The rule-base of GADIS is constructed by using a genetic algorithm for rule extraction from a patient database. Experimental results show a very good diagnostic performance in terms of accuracy, sensitivity and specificity of the intelligent system. The rule-base can be refined each time the patient database is updated over a limit (Beligiannis, et al., 2006). A second approach, that combines a fuzzy expert system with an evolutionary algorithm, concerns a hybrid intelligent system, called HIGAS, which deals with diagnosis and treatment consultation of acid-base disturbances based on blood gas analysis data. The system mainly consists of a fuzzy expert system that incorporates an evolutionary algorithm in an off-line mode. The diagnosis process, the input variables and their values were modeled based on expert’s knowledge and existing literature. The fuzzy rules are organized in groups to be able to simulate the diagnosis process. Differential Evolution (DE) algorithm is used to fine-tune the membership functions of the fuzzy variables. Medium scale experimental results show that HIGAS does better than its non-hybrid version, non-experts and other previous computer-based approaches (Koutsojannis, et al., 2006). A third approach extends the previous one by combining fuzzy neurules with DE algorithm. Fuzzy neurules are integrated rules that combine a neuro-fuzzy unit, the fuzzy adaline unit, and symbolic rules. FUNEUS is the system implementing fuzzy neurules and their inference mechanism. The deferential evolution algorithm is used to tune parameters of the membership functions for better system performance. The whole approach has been applied to data for diagnosis of heart diseases with promising results (Koutsojannis, et al., 2007).

The Biomedical Engineering Research Group in the Dept. of Information & Communication Systems Engineering - University of the Aegean (contact Person: Ilias Maglogiannis ) focuses its main research activities in the fields of medical informatics, intelligent health information systems, analysis and processing of biological data, medical image processing (segmentation, feature extraction, classification), implementation of parallel algorithms in grid infrastructures, and networked multimedia systems.

In the context of artificial intelligence and applications in medical image processing the group has conducted extended studies on image analysis (Maglogiannis, et al., 2004) and computational vision-based diagnostic systems for dermatology (Maglogiannis, et al., 2005), oncology (Makedon, et al., 2006) and microscopy (cancer imaging (Anagnostopoulos, et al., 2006), fibrosis (Maglogiannis, et al., 2008), angiogenesis (Doukas, et al., 2008) and apoptosis (Doukas & Maglogiannis, 2007). For instance, the developed skin image analysis prototype software uses extracted features from dermatological images for skin lesion classification by employing artificial intelligence methods, i.e., Discriminant Analysis, Neural Networks, and Support Vector Machines. Within the same domain of intelligent medical image processing they have developed a suite of tools for Computer Supported Angiogenesis Quantification Using Image Analysis and Statistical Averaging. A similar automated tool for apoptosis quantification is also available. Programmed cell death, also known as apoptosis, is of fundamental importance in many biological processes and also highly associated with serious diseases like cancer and HIV. The group has developed an innovative method for apoptosis phenomenon characterization based on apoptotic cell quantification and detection using active contours (snakes).

In the field of biological data processing, a parallel microarray data analysis platform has been developed. Microarray experiments produce very large amount of gene expression data and the process of statistical analysis applied to them in order to reveal the significance and correlation between some of the genes is extremely computationally intensive. In order to provide a remedy to this problem, grid computing technology was exploited. The HECTOR project (Maglogiannis, et al., 2008), is an application that deployed over the Hellenic Grid Infrastructure (HellasGRID) providing to its users an easy to use analysis tool that implements normalization and statistical selection routines, as well as gene clustering techniques (hierarchical clustering, K-means, Fuzzy C means) revealing correlations among significantly expressed genes.

Research in the area of Biomedical Informatics (BMI) is a strategic direction of the Machine Learning, Data Mining and Knowledge Discovery group at FORTH-ICS (ML-DM/KD - http://www.ics.forth.gr/bmi/data_mining.html). In fact, the group is part of the Biomedical Informatics laboratory of FORTH-ICS. BMI constitutes one from the better and more achieved examples of interdisciplinary collaboration and synergy between Medical Informatics, Bioinformatics and Biology. The aim is to support the emerging needs of individualised medicine in the raising post-genomics era. In this context, the last five years the group has focused its R&D activities on the design and development of methodologies, techniques, algorithms, tools and services to support the involved communities (molecular biologists, clinical researchers and bioinformaticians) in their daily (heavily dependent on huge amounts of genomic data) activities, and present solutions on an extended spectrum of engaged problems: (i) better ‘in-silico’ prediction of base-calls in DNA sequencing (with neural network approaches), (ii) more acurate recognition of ‘in silico’ recognition of genes in DNA sequences, (iii) discovery of diagnostic and prognostic biomarkers from microarray/gene-expression and proteomic (mass-spec) data (with advanced and novel feature selection and classification approaches), (iv) integration of heterogeneous and distributed clinico-genomic data-sources via the design, (v) development and operationalization of an Integrated Clinico-Genomics Environment (ICGE), (vi) development of combined clinico-genomic knowledge discovery workflows based on Web-Services technology, and (vii) modelling and reasoning with molecular regulatory (and metabolic) pathways. In addition, members of the ML-DM/KD group are actively involved in ERCIM’s Biomedical Informatics Working Group (Potamias, et al., 2004a; 2004b; 2006).

 

2.4 LEARNING/EDUCATION

Research work in this field is conducted at the Hellenic Open University (HOU) (contact Person: Dimitris Kalles ). More specifically the research team uses decision trees and genetic algorithms to analyze the academic performance of students who enrol in the undergraduate program on informatics throughout an academic year. Based on the accuracy of the generated rules and on knowledge of the domain, the educational impact of specific tutoring practices is analyzed and the following academic performance indices (at junior and senior years) may be used as an alert system at several levels: a) Academic performance alert for students, b) Consistency alert for a group of tutors, and c) Consistency alert for a study program.

The organization-wide adoption of such schemes is a difficult task since, besides requiring technical improvement to enhance the legitimacy of the calculations, communicating an alert to potential users poses significant political and ethical challenges. The technical infrastructure to develop performance indices based on individual student models is based on a combination of decision trees and genetic algorithms, though other techniques can be used as well (Kalles & Pierrakeas, 2006a; 2006b; Hadzilacos, et al., 2006; Hadzilacos, et al., 2008).

The Computing Systems Laboratory (Prof. G. Papakonstantinou) in the Department of Electrical and Computer Engineering NTUA, in the frame of European projects, (DEFACTO, ESPRIT-LTR 23456), (HILDE, GSRT-EPET2), has developed a system for training elementary school students in writing interactive stories (Sgouros, et al., 1996). Furthermore in the frame of the project (HILDE: Hypermedia Intelligent Learning Environment, GSRT-EPET2), a system was developed for training doctors in learning the interpretation of ECG waveforms, using an expert system cell and learning techniques (Bourlas, et al., 1996).

Artificial Intelligence Group (AI GROUP) (contact Person: Ioannis Hatzilygeroudis ) in the Laboratory of Graphics, Multimedia & GIS - Department of Computer Engineering & Informatics - University of Patras (http://mmlab.ceid.upatras.gr/aigroup/) exhibits research efforts, in the domain of e-Learning/Education, related to the use of integrated approaches in intelligent tutoring systems (ITSs). They have used a neurule-based expert system to represent pedagogical knowledge and make decisions during the teaching process in an ITS. Neurules are a type of integrated rules combining symbolic rules with neurocomputing (Hatzilygeroudis & Prentzas, 2004). Other research efforts, in the same domain, are related to the use of AI approaches, hybrid or not, in achieving adaptiveness in web-based intelligent educational systems (WBIESs), more specifically as far as adaptive testing/assessment is concerned. So, one such effort deals with achieving knowledge-based adaptive testing/assessment in a WBIES. More specifically, they focus on a mechanism for on-line creation of a user-adapted test, which can be used alongside a predetermined test. The user can ask for such a test any time he/she is willing to do so, even if he/she has not studied all predetermined concepts of a learning goal. A small rule base is used by an expert system inference engine for making decisions on the difficulty level of the exercises to be included in the test. This is based on the evaluation of the learner during concept studying. Adaptive assessment of the learner can be repeatedly used until there is no further need (Hatzilygeroudis, et al., 2006). A similar effort has resulted in using a hybrid AI approach for determination of the difficulty levels of the provided exercises. More specifically, a combination of the rule-based approach and a genetic algorithm approach is used. A genetic algorithm is used to extract some kind of rules from the data acquired from the interactions of the students with the system when answering to questions/exercises. Those rules are used to modify expert rules provided by the tutor. In this way, feedback from the students is taken into account for determination of the difficulty levels of the questions/exercises. This is important because the difficulty levels of the exercises are taken into account for the evaluation of the knowledge levels of the students with regards to various concepts. Experimental results show that a significant part of questions/exercises may need to change their level of difficulty. Furthermore, the validity of the method is experimentally showed (Koutsojannis, et al., 2007).

 

2.5 ENGINEERING AND INDUSTRY

The Machine Design Lab. of the Dept. of Mechanical Eng. and Aeronautics at the University of Patras (http://www.mech.upatras.gr/~dentsora/) has developed novel design methodologies and techniques based on advances in the fields of computational intelligence and case-based reasoning (Dentsoras, 2005; Saridakis & Dentsoras, 2006; Saridakis & Dentsoras, 2007). They applied their methods in three design cases, namely, design of vibrating conveyors, formwork design for slabs, and design of an electrostatic robotic gripper for handling fabrics. They introduced an integrated approach that organizes the design knowledge via flexible hierarchical structures and considers multiple computational intelligence techniques for implementing several design tasks in the phase of detailed design. Fuzzy logic was used in order to: a) express the fuzziness of designers’ preferences and b) extend the concept of Design Structure Matrices so that the latter may accept fuzzy preferences assigned by multiple designers and produce overall preference values. Optimum solutions were deduced via an evolutionary algorithm and/or case-based techniques. In the latter case, a neural network provided sets of retrieved solutions that converged towards current designer’s preferences.

The Computing Systems Lab. of the National Technical University of Athens has developed a system for the automatic hardware design based and on AI techniques (Economakos, et al., 2002). The complexity of modern digital systems requires complex design entry methods and thus, language based designs are often an appealing alternative for schematics. Language based design entry, supports high-level design transformations through formal and executable traditional compiler construction problem specifications, their main advantages being modularity and declarative notation. This group exploits this idea under a powerful compiler construction system and a methodology is given to design formal and executable high-level hardware manipulators. In effect, this methodology stands as a meta-level between hardware transformations and their implementation and can be valuable in fast evaluation of new ideas and techniques. Moreover, this group has developed a platform for the automatic design of hardware AI systems, e.g., PROLOG machines (Panagopoulos, et al., 2004). Conventional approaches in the implementation of logic programming applications on embedded systems are solely of software nature. As a consequence, a compiler is needed that transforms the initial declarative logic program to its equivalent procedural one, to be programmed to the microprocessor. This approach increases the complexity of the final implementation and reduces the overall system’s performance. On the contrary, presenting hardware implementations which are only capable of supporting logic programs prevents their use in applications where logic programs need to be intertwined with traditional procedural ones, for a specific application. This group exploits HW/SW codesign methods to present a microprocessor, capable of supporting hybrid applications using both programming approaches. They take advantage of the close relationship between attribute grammar evaluation and knowledge engineering methods to present a programmable hardware parser that performs logic derivations and combine it with an extension of a conventional RISC microprocessor that performs the unification process to report the success or failure of those derivations. The extended RISC microprocessor is still capable of executing conventional procedural programs, thus hybrid applications can be implemented. The presented implementation is programmable, supports the execution of hybrid applications, increases the performance of logic derivations (experimental analysis yields an approximate 1000% increase in performance) and reduces the complexity of the final implemented code.

The Laboratory for Automation and Robotic of the University of Patras in collaboration with the Dept. of Informatics and Telecommunications Technology TEI of Epirus have conducted research on controlling industrial processes (Papageorgiou, et al., 2002; Papageorgiou, et al., 2006). To this end, they use Fuzzy Cognitive Maps (FCMs), an attractive knowledge-based methodology combining the robust properties of fuzzy logic and neural networks. FCMs represent causal knowledge as a signed directed graph with feedback and provide an intuitive framework which incorporates the experts’ knowledge. FCMs handle available information and knowledge from an abstract point of view. They develop behavioural model of the system exploiting the experience and knowledge of experts. The construction of FCMs is based mainly on experts who determine the structure of FCM, i.e. concepts and weighted interconnections among concepts. But this methodology may not be a sufficient model of the system because the human factor is not always reliable. Thus the FCM model of the system may require restructuring which is achieved through adjustment the weights of FCM interconnections using specific learning algorithms for FCMs. To handle this problem, they use two unsupervised learning algorithms, namely, active Hebbian learning and nonlinear Hebbian learning for training FCMs. This proposed learning procedure is a promising approach for exploiting experts’ involvement with their subjective reasoning and at the same time improving the effectiveness of the FCM operation mode and thus it broadens the applicability of FCMs modeling for complex systems.

 

2.6 TELECOMMUNICATIONS AND WEB

The Logic Programming and Intelligent Systems Group (http://lpis.csd.auth.gr) of the Aristotle University was involved in the development of ExperNet (Vlahavas, et al., 2002), a multi-agent knowledge-based system providing intelligent management services to network administrators, with the aim to increase the functionality of the current management systems and provide the means to deliver higher quality services to the end user. ExperNet offers constant monitoring of the state of its target network, including both the critical parts, such as routers and leased lines, and common network services, such as FTP or HTTP. ExperNet assists network operators to detect and diagnose hardware failure and network traffic problems and suggests the most viable at the time solution. These error reporting and diagnosis capabilities can significantly decrease the downtime of the network components, leading to an increased availability of the overall network and support further the administrator by offering immediate expert suggestions about repair actions that should be taken in order to resolve the abnormal network situations. It is based on existing and widely used network management protocols (SNMPv2), which makes its application to any existing TCP/IP network possible. Managing a WAN is inherently a distributed task and this inevitably led to modeling ExperNet as a multi-agent system. At each network management node there is one agent, covering the area that the node is responsible for. Consequently, the structure of the current system architecture goes in line with the structure of the pre-existing organization of the experimental zone of the Ukrainian national network, but it can be easily adapted to closely reflect the organizational structure any WAN. Each ExperNet agent is able to co-operate with other agents towards problem diagnosis and repair, through the use of social knowledge for co-ordination. Agents have also to communicate with network management software, to acquire an accurate picture of the network state in terms of device installation, removal, reachability and operational parameters. Consequently, the development of ExperNet involved the development/extension and integration of a set of advanced tools each dedicated to a specific task, ie. the CS-Prolog II logic programming platform that addressed integration and communication requirements, the DEVICE knowledge base system that addressed agent reasoning requirements and the HNMS+ and BigBrother that addressed accurate network monitoring requirements with a low bandwidth cost. These tools performed well in a real network environment and some were further extended after the project's completion (Bassiliades, et al., 2000; Sakellariou, et al., 2006). The ExperNet system has been installed and tested in a real network environment in Ukraine and has performed well monitoring a network of significant size. As we saw from the performed tests, ExperNet showed robustness on a set of typical management cases that network operators of the Ukrainian national network meet in real, every-day practice. As a prototype it clearly demonstrated the benefits of applying AI technology to network management.

The Logic Programming and Intelligent Systems research group of the Aristotle University of Thessaloniki has implemented a system called PersoNews (http//news.csd.auth.gr), a machine learning enhanced adaptive RSS news reader, in order to alleviate the WWW information overload problem (Katakis, et al., 2008). The main features of the PersoNews system are: a) the aggregation of many different news sources that offer an RSS version of their content, b) incremental filtering offering dynamic personalization of the content not only per user but also per each feed a user is subscribed to, and c) the ability for every user to watch an abstract topic of interest by keyword-based filtering through a taxonomy of topics. PersoNews performs periodical polling of all feeds that it monitors in order to retrieve new publications and store meta-data like title, description, date and URL in the system’s database. At the same time, it performs content filtering by classifying new publications into interesting/junk for each user according to their interests. PersoNews offers two channels of personalization: Feed filtering and topic filtering. In each case, filtering is achieved by using machine learning technology. Feed filtering personalizes the dissemination of news of a specific feed to each specific user that chooses to monitor it. The main goal of the filtering mechanism is the automatic discrimination of incoming articles into interesting and uninteresting ones according to user preferences. The machine learning element in PersoNews necessary for filtering uninteresting articles comprises two basic components: a) an incremental feature based classifier, and b) an incremental feature ranking method. Feature-based classifiers are those classifiers that can consider any subset of features for the classification of a new instance. Two inherently feature-based algorithms are Naïve Bayes and k-Nearest Neighbourhood. Incremental feature ranking methods, such as information theoretic (Information Gain, Chi-Squared Measure) term selecting functions can deal with a dynamic feature space, as they calculate statistics for each feature independently. The chi-squared method was selected as the feature ranking component of the proposed framework, due to its simplicity and effectiveness in text categorization problems. The Naïve Bayes algorithm was selected for instantiating the learning module of the proposed framework.

The Artificial Intelligence Group of the Wire Communications Laboratory at the University of Patras (http://www.wcl.ece.upatras.gr/ai) has been involved in a series of EU-funded projects for incorporating speech understanding technology to telecommunication systems (Kostoulas, et al., 2008; Mporas, et al., 2008). The VASME project aimed at the design and development of a voice operated VHF automatic on demand information system for ships and vessels. The major purpose of this system is to increase the flow of information exchange between vessels and traffic control towers. It provides real-time access to a set of value added services including administrative information, commercial information, technical information and navigational support. This system is based on speaker-independent speech recognition, noise reduction, voice synthesis, relational data bases, and real time systems with high computational charge. The ACCeSS project addresses a first step for automation of call centers for highly personal intensive applications in insurance companies. Several insurance operations are automated using the telephone or post mail for the contact between an insurance company and its customers. The main objectives of the project are twofold: First the development of reliably working speech dialogue systems based on speech understanding and speech generation technology including intensive contact with databases. Second, a text understanding system will be developed to extract the relevant information from form sheets, returning a very large number after promotional action. The IDAS project (Lehtinen, et al., 2000) addresses the challenging problem of automating the provision of directory assistance services to the public over the telephone network. The technical challenge tackled by this project makes high demands on each of the speech processing components: a speech recognizer that can distinguish the uttered word out of a large vocabulary, independently of the speaker’s voice and the mostly poor signal quality, a speech production system able to speak out any imaginable phone directory entry (containing names and words from different languages), a dialogue component that can interpret user inputs and to ask the right questions in order to guide the users quickly to their desired information and out of misunderstandings. The MoveON project investigates the application of multi-modal and multi-sensor zero-distraction interfaces for enabling 2-wheel vehicle drivers to access online, in real-time and taking into account on the road safety issues, services and information resources. MoveON also supports the complimentary use of speech, head nods and tactile modalities pushing beyond the state-of-the-art for motorcycle set-ups. The project target users are the police force motorcyclists and motorcycle riders in all walks of life. The scientific objectives address the development of a robust speech-processing module, which is complemented with a dialogue system incorporating speech recognition and a text-to-speech component. MoveON is deploying the Usability Relationship Evaluation Methodology on controlled field-experiments for assessing the user's distraction while interacting with the prototype system. The project develops evaluation scenarios that energize situations in which to validate the motorcyclists’ driving attention disruption while interacting with the MoveON multimodal interface.

 

2.7 PERVASIVE SYSTEMS

The Biomedical Engineering Research Group in the Dept. of Information & Communication Systems Engineering - University of the Aegean (contact Person: Ilias Maglogiannis ) has done research in intelligent pervasive applications and the corresponding enabling technologies. The details of their work regarding pervasive healthcare systems in either controlled environments (e.g., health care units or a hospitals), or in sites where immediate health support is not possible (i.e. the patient’s home or an urban area) can be found in (Doukas & Maglogiannis, 2007). Special focus is raised on intelligent platforms (e.g., agents, context-aware and location-based services, and classification systems) that enable advanced monitoring and interpretation of patient status and environment optimizing the whole medical assessment procedure. Within this concept and an initial implementation of a patient status awareness system that may be used for patient activity interpretation and emergency recognition in cases like elder falls and distress speech expressions has been developed (Doukas & Maglogiannis, 2008). The awareness is performed through collecting, analyzing and classifying motion and sound data. The latter are collected through sensors equipped with accelerometers and microphones that are attached on the body of the patients and transmit patient movement and sound data wirelessly to the monitoring unit. Applying Short Time Fourier Transform (STFT) and spectrogram analysis on sounds detection of fall incidents is possible. The classification of the sound and movement data is performed using Support Vector Machines. Evaluation results indicate the high accuracy and the effectiveness of the proposed implementation. The system architecture is open and can be easily enhanced to include patient awareness based on additional context (e.g., physiological data).

Dr. Yannakakis at the IT-University of Copenhagen, Denmark (http://www.itu.dk/~yannakakis/), has research at the field. Among his theoretical contributions, the most important are the establishment of generic cognitive and affective models of entertainment in specific genres of computer games; a scheme for obtaining digital entertainment of richer interactivity and higher enjoyment; and the design of robust controllers, that successfully drive multiple agents with limited inter-communication, using minimal effort to construct them. Pursuing this notion of player satisfaction, Dr. Yannakakis defined a generic quantitative operational measure of player satisfaction, called `interest', for simple prey/predator games. This measure was based on psychological studies, inspired by previous work on entertainment estimates for chess and empirically established as an efficient and reliable entertainment metric corresponding well to human judgment for a test-bed game platform (Yannakakis & Hallam, 2007). Further studies have shown that machine learning can extract a better estimator of player satisfaction than a custom-designed (or designer-driven) one, given appropriate estimators of entertainment factors of the game and data on human players' preferences. Applying this effective methodology developed in simple prey/predator games, successful models of player satisfaction were constructed for simple physical interactive games designed on the “Playware” augmented-reality platform. Models were built on features derived from both player-platform interaction and physiological indices (Yannakakis, et al., 2008; Yannakakis & Hallam, 2008).

A robust on-line neuro-evolution learning mechanism, based on the above entertainment models, was shown to be capable of maintaining or increasing the game's entertainment value while the game was being played (Yannakakis & Hallam, 2007). Furthermore, studies with the ``Playware'' playground have shown that real-time adaptive ad-hoc rule-based mechanisms may improve children's gameplay experience in physical interactive playgrounds.

Supported by FTP (Danish Research Council) project no: 274-05-0511, Dr. Yannakakis has established the field of player satisfaction modeling through publications in top-ranked journals and organization of two specialized international workshops on “Optimizing Player Satisfaction”. Moreover, Dr. Yannakakis chairs the task force of IEEE-Computational Intelligence Society on player satisfaction modeling and has been invited by AAAI press to write the introductory book in the field of optimizing player satisfaction (Yannakakis & Hallam, 2009).

 

3. MATHEMATICAL FOUNDATIONS

The Laboratory for Automation and Robotic of the University of Patras (http://www.lar.ee.upatras.gr/) in collaboration with the Dept. of Informatics and Telecommunications Technology TEI of Epirus have conducted research on Fuzzy Cognitive Maps (FCMs), a modeling methodology for complex systems originating from the combination of Fuzzy Logic and Neural Networks. An FCM describes the behavior of a system in terms of concepts; each concept represents an entity, a state, a variable, or a characteristic of the system. FCMs are developed by human experts - who operate/supervise the system in such a way that the accumulated experience and knowledge are integrated in a causal relationship among factors/components of the system. Experts involved in the construction of FCM determine concepts and causality among them. This approach may yield to a distorted model, because experts may not consider the appropriate factors and they may have assigned inappropriate causality weights among FCM concepts. Weight adaptation methods are very promising as they can alleviate these problems by allowing the creation of less error prone FCMs where causal links are adjusted through a learning process. The learning procedure is a technique which increases the efficiency and robustness of FCMs. Moreover, the learning rules supply FCMs with useful characteristics such as the ability to learn arbitrary nonlinear mappings and capability to generalize situations (Papageorgiou, et al., 2004). The Laboratory for Automation and Robotic group proposed the NHL algorithm based on the nonlinear Hebbian-type unsupervised learning rule (Papageorgiou, et al., 2003; Papageorgiou & Groumpos, 2005), which has been modified and adapted for FCMs. The NHL algorithm is based on the premise that all the concepts in FCM model are synchronously triggering at each iteration step and change their values synchronously. During this triggering process all weights of the causal interconnections of the concepts are updated and a modified weight is derived. The proposed learning algorithm extracts hidden and valuable knowledge of experts and it can increase the effectiveness of FCMs.

Furthermore, the group has proposed Active Hebbian Algorithm (AHL) (Papageorgiou, et al., 2004b), an advanced method for learning FCM based on the Hebbian algorithm. In AHL, experts also identify the sequence of activation concepts in the FCM model. The novelty of this algorithm is based on introducing the sequence of influence from one concept to another; in this way, the interaction cycle is dividing in steps. When the experts develop the FCM, they are asked to determine the sequence of activation concepts, the activation steps and the activation cycle. At every activation step, one (or more) concept(s) becomes Activated concept(s), triggering the other interconnecting concepts, and in turn, at the next simulation step, may become Activation concept. When all the concepts have become Activated concepts, the simulation cycle has closed and a new one starts until the FCM model converges in an equilibrium region. An activation cycle consists of steps; at each activation step one or more concepts are the Activation concepts that influence the interconnected concepts until the termination of the sequence of activation closes the cycle. In addition to the determination of sequence of activation concepts; experts select a limited number of concepts as outputs for each specific problem which are defined as the Activation Decision Concepts (ADCs). These concepts are in the center of interest; they stand for the main factors and characteristics of the system, known as outputs and their values represent the final state of the system (Papageorgiou, et al., 2006).

The AI Research Group at the Department of Computer Science, University of Ioannina, focuses on the area of solving differential equations using artificial neural networks with applications in many domains of Engineering and Science. More specifically, the general problem of solving differential equations has been formulated as a supervised learning problem, which is solved by training feed-forward neural networks using an appropriately defined error function (Lagaris, et al., 2000). The method has been proved efficient and accurate and also exhibits the potential of direct hardware implementation.

The Artificial Intelligence Group of the Wire Communications Laboratory at the University of Patras (http://www.wcl.ece.upatras.gr/ai) has been involved in research on theoretical and mathematical models of artificial intelligence methods and algorithms, such as search methods, problem solving, rule based systems, knowledge representation, logic programming, machine learning, intelligent human-machine interaction, user modeling, automata theory (Sgarbas, et al., 2003), game playing, and quantum artificial intelligence (Sgarbas, 2007).

The Intelligence, Modelling & Computation (IMC) research group at CITY College (http://www.city.academic.gr/special/research/imc/index.html) is engaged with research in mathematical foundations and formal aspects of modelling agents and multi-agent systems. More specifically, the group is researching Formal Modelling, Testing & Verification (X-machines, Communicating X-Machines) of Intelligent Agent and Multi-Agent Systems, as a continuation of the group's previous research in formal methods and in particular in formal modeling, testing and verification of complex systems. The group focuses on formal modeling of such systems, and in particular they investigate how state-based formal methods, such as X-machines and its Communicating counterpart could facilitate modeling of agents and multi-agents systems (Kefalas, et al., 2003; Eleftherakis, et al., 2003). The rationale behind that was the ability to apply formal testing and verification through model checking towards the implementation of correct agent systems. A formal framework, including a lightweight methodology for the development of such systems was developed. This included a language for the formal description of agents as well as the formal foundations of complete testing and model checking accompanied by several tools. The IMC group is in close collaboration with the Verification & Testing research group of the Department of Computer Science of the University of Sheffield, through the South-East European Research Center (SEERC).

 

4. CLOSURE AND REMARKS

It is evident that the AI community in Greece is active in developing a wide area of applications covering areas such as Artificial Life, Business, Commerce, Finance, Government, Health, Education, Engineering, Industry, Telecommunications, Web, and Pervasive Systems. It has to be underlined that the reported areas as well as the works and groups in each application area are only indicative. Moreover, a noticeable contribution of Greek researchers towards the establishment of mathematical foundations has been reported.

An important conclusion that can be drawn by this survey is that the vast majority of AI research in Greece is conducted by academic institutions and in the framework of collaborative research projects. The Greek industries are not significantly involved in the developing and/or exploitation of AI applications. As a result, many attempts to build intelligent systems stop at the production of prototypes rather than commercial products. Therefore, a closer collaboration of Greek research institutions with industry is a great challenge for the years to come. To this end, the Greek Government, through the General Secretariat of research and Technology, should further enforce funding of mixed Academic-Industrial projects on issues related to the development of Artificial Intelligence Applications and their penetration to the industrial-business world, as well as the Greek society, in general.

 

REFERENCES

Anagnostopoulos, I., and Maglogiannis, I. (2006) Neural network-based diagnostic and prognostic estimations in breast cancer microscopic instances. Medical & Biological Engineering & Computing, 44(9), pp. 773-784.

Bassiliades, N., Vlahavas, I., and Elmagarmid, A. (2000) E-DEVICE: An Extensible Knowledge Base System with Multiple Rule Support. IEEE Transactions on Knowledge and Data Engineering, 12(5), pp. 824-844.

Beligiannis, G., Hatzilygeroudis, I., Koutsojannis, C., and Prentzas, J. (2006) A GA Driven Intelligent System for Medical Diagnosis, in B. Garbys, R. J. Howlett and L. C. Jain (Eds), Proc. 10th International Conference on Knowledge-Based Intelligent Information and Engineering Systems (KES-2006), Bournemouth, UK, October 2006, Part I, LNAI 4251, Springer, pp. 968-975.

Blekas, K., Fotiadis, D. I., and Likas, A. (2003) Greedy Mixture Learning for Multiple Motif Discovery in Biological Sequences. Bioinformatics, 19(5), pp. 607-617.

Blekas, K., Fotiadis, D. I., and Likas, A. (2005) Motif-Based Protein Sequence Classification Using Neural Networks. Journal of Computational Biology, 12(1), pp. 64-82.

Blekas, K., Galatsanos, N., Likas, A., and Lagaris I. E. (2005) Mixture Model Analysis for DNA Microarray Images. IEEE Trans. on Medical Imaging, 24(7), pp. 901-909.

Bourlas, F., Giakoumakis, E., Koutsouris, D., Papakonstantinou, G., and Tsanakas, P. (1996) The CARDIO-LOGOS system for ECG training and diagnosis. Journal of Technology and Health Care, 3, pp 279-285.

Dentsoras, A.J. (2005) Information Generation during Design: Information Importance and Design Effort. Artificial Intelligence in Engineering, Design, Analysis and Manufacturing, 19, pp. 19-32.

Doukas, C., and Maglogiannis, I. (2007) Automated Cell Apoptosis Characterization using Active Contours. Conf Proc IEEE Eng Med Biol Soc., (1), pp. 812-815.

Doukas, C., and Maglogiannis, I. (2008) Intelligent Pervasive Healthcare Systems. Advanced Computational Intelligence Paradigms in Healthcare. Studies in Computational Intelligence, 107, Sordo, Margarita; Vaidya, Sachin; Jain, Lakhmi C. (Eds.).

Doukas, C., Maglogiannis, I., and Chatziioannou, A. (2008) Computer Supported Angiogenesis Quantification Using Image Analysis and Statistical Averaging. IEEE Transactions on Information Technology in Biomedicine, accepted for publication.

Economakos, G., Economakos, P., Poulakis, I., Papakonstantinou, G., and Georgoulis, S. (2002) Handling Advanced Scheduling Heuristics under a Hardware Compiler Generation Environment. Knowledge Based Systems, 15.

Eleftherakis, G., Kefalas, P., and Sotiriadou, A. (2003) Formal Verification of Reactive Agents for Intelligent Control. Proceedings of the 12th ISAP Intelligent System Applications to Power Systems Conference.

Hadzilacos, Th., Kalles, D., and Pierrakeas, C. (2008) On Developing and Communicating User Models for Distance Learning based on Assignment and Exam Data, (to appear in) Intelligent Interactive Systems in Knowledge-based Environments, M. Virvou and L. Jain (eds), Springer.

Hadzilacos, Th., Kalles, D., Pierrakeas, C., and Xenos, M. (2006) On Small Data Sets revealing Big Differences. Πανελλήνιο Συνέδριο Τεχνητής Νοημοσύνης, Ηράκλειο.

Hatzilygeroudis, I., and Prentzas, J. (2004) Using a Hybrid Rule-Based Approach in Developing an Intelligent Tutoring System with Knowledge Acquisition and Update Capabilities. Journal of Expert Systems with Applications, 26(4), pp. 477-492.

Hatzilygeroudis, Ι., Koutsojannis, C., Papavlasopoulos, C., and Prentzas, J. (2006) Knowledge-Based Adaptive Assessment in a Web-Based Intelligent Educational System, Proceedings of the 6th IEEE International Conference on Advanced Learning Technology (ICALT-2006), pp. 651-655.

Kalles, D., and Pierrakeas, C. (2006) Analyzing Student Performance in Distance Learning with Genetic Algorithms and Decision Trees. Applied Artificial Intelligence, 20(8), pp. 655–674.

Kalles, D., and Pierrakeas, C. (2006) Using Genetic Algorithms and Decision Trees for a posteriori Analysis and Evaluation of Tutoring Practices based on Student Failure Models. 3rd IFIP Conference on Artificial Intelligence Applications and Innovations, Athens, Greece.

Katakis, I., Tsoumakas, G., Banos, E., Bassiliades, N., and Vlahavas, I. (2008) An Adaptive Personalized News Dissemination System. Journal of Intelligent Information Systems, Springer.

Katevas, N, Sgouros, N. M., Tzafestas, S., Papakonstantinou, G., Beattie, P., Bishop, J. M., Tsanakas, P., and Koutsouris, D. (1997) SENARIO: A sensor-aided intelligent navigation system for powered wheelchairs. IEEE Robotics and Automation Magazine, 4(4), pp 60-70.

Kefalas, P., Eleftherakis, G., Holcombe, M., and Stamatopoulou, I. (2005) Formal Modelling of the Dynamic Behaviour of Biology-Inspired Agent-based Systems. Molecular Computational Models: Unconventional Approaches. Gheorghe M. (Ed.), Idea Group Publishing, pp. 243-276.

Kefalas, P., Holcombe, M., Eleftherakis, G., and Gheorge, M. (2003) A Formal Method for the Development of Agent Based Systems. Intelligent Agent Software Engineering. Plekhanova V. (Ed.), Idea Group Publishing, pp.68-98.

Kostoulas, T. Mporas, I., Ganchev, T., Katsaounos, N., Lazaridis, A., Ntalampiras, S., and Fakotakis, N. (2008) LOGOS: A Multimodal Dialogue System for Controlling Smart Appliances. In Proceedings of the 1st International Symposium on Intelligent Interactive Multimedia Systems and Services.

Koutsojannis, C., and Hatzilygeroudis, I. (2006) Fuzzy-Evolutionary Synergism in an Intelligent Medical Diagnosis System, in B. Garbys, R. J. Howlett and L. C. Jain (Eds), Proceedings of the 10th International Conference on Knowledge-Based Intelligent Information and Engineering Systems (KES-2006), Bournemouth, Part II, LNAI 4252, Springer, 1313-1322.

Koutsojannis, C., and Hatzilygeroudis, I. (2007) Using a Neurofuzzy Approach in a Medical Application, in Bruno Apolloni, Robert J. Howlett and Lakhmi Jain (Eds), Proceedings of the 11th International Conference on Knowledge-Based Intelligent Information and Engineering Systems (KES-2007), Vietri sul Mare, Italy, LNAI 4693, Springer, 477-484.
Koutsojannis, C., Beligiannis, G., Hatzilygeroudis, Ι., Papavlasopoulos, C., and Prentzas, J. (2007) Using a hybrid AI approach for exercise difficulty level adaptation. International Journal of Continuing Engineering Education and Life-Long Learning (IJCEELL), Special Issue on “Integrating Intelligent and Adaptive Hypermedia Techniques in Web-Based Education Systems”, 17(4-5), pp. 256-272.

Lagaris, I. E., Likas, A. and Papageorgiou, D. G. (2000) Neural Network Methods for Boundary Value Problems with Irregular Boundaries. IEEE Transactions on Neural Networks. 11(5), pp. 1041-1049.

Lehtinen, G., Safra,S., Gauger, M., Cochard, J.-L,.Kaspar, B., Hennecke, M.E., Pardo, J.M. Cordoba, R., San-Segundo, R., Tsopanoglou, A., Louloudis, D., and Mantakas, M. (2000) IDAS: Interactive Directory Assistance Service. In Proceedings of the International Workshop on Voice Operated Telecom Services, pp. 51–54.

Maglogiannis, I. and Zafiropoulos, E. (2004) Utilizing Support Vector Machines for the Characterization of Digital Medical Images. BMC Medical Informatics and Decision Making, 4(4).

Maglogiannis, I., Pavlopoulos, S., and Koutsouris, D. (2005) An Integrated Computer Supported Acquisition, Handling and Characterization System for Pigmented Skin Lesions in Dermatological Images. IEEE Transactions on Information Technology in Biomedicine, 9(1), pp 86-98.

Maglogiannis, I., Sarimveis, H., Kiranoudis, C., Chatziioannou, A. A., Oikonomou, N., and Aidinis ,V. (2008) Radial Basis Function neural networks classification for the recognition of idiopathic pulmonary fibrosis in microscopic images. IEEE Transactions on Information Technology in Biomedicine, 12(1), pp. 42–54.

Maglogiannis, I., Soldatos, J., Chatziioannou, A., Milonakis, V., and Kanaris, Y. (2008) An Application Platform Enabling High Performance Grid Processing of Microarray Experiments. 20th IEEE International Symposium on Computer-Based Medical Systems.

Makedon, F., Karkaletsis, V., and Maglogiannis, I. (2006) Computational Analysis and Decision Support Systems in Oncology. Guest Editorial Overview. Oncology Reports special issue “Computational Analysis and Decision Support Systems in Oncology”, 15, pp. 971-974.

Mporas, I., Ganchev, T., and Fakotakis, N. (2008) A Hybrid Architecture for Automatic Segmentation of Speech Waveforms, Proc. of the International Conference on Acoustics Speech and Signal Processing.

Panagopoulos, I., Pavlatos, C., and Papakonstantinou, G. (2004) An Embedded System for Artificial Intelligence Applications. International Journal of Computational Intelligence, 1(1).

Papageorgiou, E. I., and Groumpos, P. P. (2005) A weight adaptation method for fine-tuning Fuzzy Cognitive Map causal links. Soft Computing Journal. 9, pp. 846-857.

Papageorgiou, E. I., Stylios C. D., and Groumpos, P. P. (2002) Activation Hebbian Learning Rule for Fuzzy Cognitive Maps. Proceedings of 15th IFAC International Federation of Automatic Control World Congress.

Papageorgiou, E. I., Stylios, C. D., and Groumpos, P. P. (2003) Fuzzy Cognitive Map Learning based on Nonlinear Hebbian Rule. Proc. 16th Australian Joint Conference on Artificial Intelligence, Perth-Western Australia. Gedeon, T. D., and Fung, L. C. C. (Eds.). LNAI 2903, Springer-Verlag, pp. 254-266.

Papageorgiou, E. I., Stylios, C. D., and Groumpos, P. P. (2004) Active Hebbian Learning to Train Fuzzy Cognitive Maps. International Journal of Approximate Reasoning. 37, pp. 219-249.

Papageorgiou, E. I., Stylios, C. D., and Groumpos, P. P. (2004) The Challenge of Using Unsupervised Learning Algorithms for Fuzzy Cognitive Maps. Proceedings of IEEE Int. Joint Conference on Neural Networks (IJCNN 2004). Budapest.

Papageorgiou, E. I., Stylios, C. D., and Groumpos, P. P. (2006) Unsupervised Learning Techniques for Fine-tuning Fuzzy Cognitive Map Causal Links. International Journal of Human-Computer Studies, 64, pp. 727-743.

Potamias, G., Koumakis, L., and Moustakis, V. (2004) Gene Selection via Discretized Gene-Expression Profiles and Greedy Feature-Elimination. LNAI 3025, Springer, pp. 256-266.

Potamias, G., Koumakis, L., and Moustakis, V. (2004) Mining XML Clinical Data: The HealthObs System. Ingenierie des systems d'information, 10(1), pp. 59-79.

Potamias, G., May, M., and Ruping, S. (2006) Grid-based Knowledge Discovery in Clinico-Genomic Data. LNBI 4345, Springer, pp. 219-230.

Sakellariou, I., Vlahavas, I., Futo, I., Pasztor, Z., and Szeredi, J., (2006) Communicating Sequential Processes for Distributed Constraint Satisfaction. Information Sciences, 176(5), pp 490-521.

Saridakis, K.M. Dentsoras, A.J. (2006) Integration of Fuzzy Logic, Genetic Algorithms and Neural Networks in Collaborative Parametric Design, 20, pp. 379–399.

Saridakis, K.M. Dentsoras, A.J. (2007) Case-DeSC: A System for Case-Based Design with Soft Computing Techniques. Expert Systems with Applications, 32, pp. 641–657.

Sgarbas, K. (2007) The Road to Quantum Artificial Intelligence. Current Trends in Informatics, Proc. PCI-2007, 11th Panhellenic Conference in Informatics, Vol. A, pp.469-477, Patras, Greece.

Sgarbas, K., Fakotakis, N., and Kokkinakis, G. (2003) Optimal Insertion in Deterministic DAWGs. Theoretical Computer Science. 301(1-3), pp.103-117.

Stamatopoulou, I., Kefalas, P., and Gheorghe, M. (2007) Modelling the dynamic structure of biological state-based systems. BioSystems, 87(2-3), pp. 142-149.

Stamatopoulou, I., Kefalas, P., and Gheorghe, M. (2008) OPERAS: a formal framework for multi-agent systems and its application to swarm-based systems. LNAI, Artikis, A., O'Hare, G., Stathis, K., and Vouros, G. (Eds.), Springer.

Tambouratzis, G., Papakonstantinou, G., Stamatelopoulos, S., Zakopoulos, N., and Moulopoulos, S. (2002) Analyzing the 24-Hour Blood Pressure and Heart-Rate Variability with Self-Organizing Feature Maps. International Journal of Intelligent Systems, 17.

Thomaidis, N. S. (2006) The implications of behavioral finance to the modeling of securities prices. Behavioral Finance. Satish, D., and Krishna Kishore, P. (Eds.). Finance Series. The ICFAI University Press.

Thomaidis, N. S. (2007) New Trends in Financial Engineering: Combining Stochastic and Computational Intelligent Methodologies. PhD Thesis. University of the Aegean, Department of Financial and Management Engineering, Chios, Greece.

Thomaidis, N. S., and Dounias, G. (2006) Cointegration and Error Correction Models: towards a reconciliation between behavioural finance and econometrics. The ICFAI Journal of Behavioural Finance. 3(3), pp. 51-73.

Thomaidis, N. S., Angelidis, T., Vassiliadis, V., and Dounias, G. (2008) Active Portfolio Management with Cardinality Constraints: An application of particle swarm optimization, New Mathematics and Natural Computation, accepted for publication.

Thomaidis, N. S., Kondakis, N., and Dounias, G. (2006) An Intelligent Statistical Arbitrage Trading System. LNAI 3955, Springer-Verlag.

Thomaidis, N. S., Nikitakos, N., and Dounias, G. (2006) The evaluation of Information Technology Projects: a Fuzzy Multicriteria Decision Making Approach. International Journal of Information Technology and Decision Making. 5(1).

Thomaidis, N. S., Tzastoudis, V., and Dounias, G. (2006) A comparison of neural network model-selection strategies for the pricing of S&P 500 stock index options. International Journal of Artificial Intelligence Tools.

Tsadiras, Α. Κ. (2005) Simulating Fuzzy Cognitive Map Models for Making Predictions. WSEAS Transactions on Information Science and Applications. 2, pp.1689-1696.

Tsadiras, A. K. (2007) Computer based Management Decisions, Based on Knowledge of Experts. Proceedings of eRA–2, Conference for the contribution of Information Technology, to Science, Economy, Society and Education. Piraeus, Greece.

Tsadiras, A. K., and Kouskouvelis, I. (2005) Using Fuzzy Cognitive Maps as a Decision Support System for Political Decisions: The Case of Turkey’s Integration into the European Union. Proc. 10th Panhellenic Conference on Informatics (PCI’ 2005), Volos, Greece, pp. 371-381.

Tsadiras, Α. Κ., and Margaritis, K. G. (1997) Cognitive Mapping and Certainty Neuron Fuzzy Cognitive Maps. Information Sciences. 101, pp.109-130.

Tsadiras, Α. Κ., Margaritis, K. G., and Mertzios, B. G. (1995) Strategic Planning Using Fuzzy Cognitive Maps. Studies in Informatics and Control. 4, pp.237-245.

Vlahavas, I., Bassiliades, N., Sakellariou, I., Molina, M., Ossowski, S., Futo, I., Pasztor, Z., Szeredi, J., Velbitskiyi, I., Yershov, S., and Netesin, I. (2002) ExperNet: An Intelligent Multi-Agent System for WAN Management. IEEE Intelligent Systems, 17(1), pp. 62-72.

Yannakakis, G. N., and Hallam, J. (2007) Towards Optimizing Entertainment in Computer Games. Applied Artificial Intelligence, 21, pp. 933-971.

Yannakakis, G. N., and Hallam, J. (2008) Entertainment Modeling through Physiology in Physical Play. International Journal of Human-Computer Studies (to appear).

Yannakakis, G. N., and Hallam, J. (2009) Modeling and Optimizing Player Satisfaction in Games: An Introduction. Invitation by AAAI/MIT Press (status: monograph proposal accepted).

Yannakakis, G. N., Hallam, J., and Lund, H. H. (2008) Entertainment Capture through Heart Rate Activity in Physical Interactive Playgrounds User Modeling and User-Adapted Interaction. Special Issue on Affective Modeling and Adaptation, 18(1-2), pp. 207-243.

Yannakakis, G. N., Levine, J., and Hallam, J. (2007) Emerging Cooperation with Minimal Effort. Rewarding over Mimicking. IEEE Transactions on Evolutionary Computation, 11(3), pp. 382-396.

Yannakakis, G. N., Levine, J., and Hallam, J. (2007) Emerging Cooperation with Minimal Effort. Rewarding over Mimicking. IEEE Transactions on Evolutionary Computation. 11(3), pp. 382-396.