Contents of Volume 17 (2007)

5/2007 4/2007 3/2007 2/2007 1/2007


  • [1] Editorial, 505-506.
  • [2] Peters S., Koenig A. (Germany): A hybrid texture analysis system based on non-linear & oriented kernels, particle swarm optimization, and kNN vs. support vector machines, 507-527.

    The presented work reports on the progress of our methodology and framework for automated image processing and analysis systém design for industrial vision application. We focus on the important task of automated texture analysis, which is an essential component of automated quality-control systems. In this context, the portfolio of texture operators and assessment methods has been enlarged. Optimized operator parameterization is investigated using particle swarm optimization (PSO). A particular goal of this work is the investigation of support vector machines (SVM) as alternative assessment method for the operator parameter optimization, incorporating the efficient inclusion of SVM parameter settings in this optimization. Methods of the enhanced portfolio were applied employing benchmark textures, real application data from leather inspection, and synthetic textures including defects, specially designed to industrial needs. The key results of our work are that SVM is a highly esteemed and powerful assessment and classification method and parameter optimization, based, e.g., on SVM/PSO of standard and proprietary texture operators boosted performance in all cases. However, the appropriateness of a certain operator proved to be highly data-dependent, which advocates our methodology even more. Thus, the operator selection has been included and investigated for the synthetic textures. Summarising, our work provides a generic texture analysis system, even for unskilled users, that is automatically configured to the application. The method portfolio will be enlarged in future work.

  • [3] Basile P., de Gemmis M., Gentile A. L., Iaquinta L., Lops P., Semeraro G. (Italy): An electronic performance support system based on a hybrid content-collaborative recommender system, 529-541.

    An Electronic Performance Support System (EPSS) introduces challenges on contextualized and personalized information delivery. Recommender systems aim at delivering and suggesting relevant information according to users preferences, thus EPSSs could take advantage of the recommendation algorithms that have the effect of guiding users in a large space of possible options. The JUMP project (JUst-in-tiMe Performance support systém for dynamic organizations, co-funded by POR Puglia 2000-2006 - Mis. 3.13, Sostegno agli Investimenti in Ricerca Industriale, Sviluppo Precompetitivo e Trasferimento Tecnologico) aims at integrating an EPSS with a hybrid recommender system.

    Collaborative and content-based filtering are the recommendation techniques most widely adopted to date. The main contribution of this paper is a content-collaborative hybrid recommender which computes similarities between users relying on their content-based profiles in which user preferences are stored, instead of comparing their rating styles. A distinctive feature of our systém is that a statistical model of the user interests is obtained by machine learning techniques integrated with linguistic knowledge contained in WordNet. This model, named ``semantic user profile'', is exploited by the hybrid recommender in the neighborhood formation process.

  • [4] Alexandrino J. L.; Zanchettin C., Carvalho Filho E. C. B. (Brazil): Artificial immune system with ART memory Hibridization, 543-554.

    The present work proposes the architecture Clonart (Clonal Adaptive Resonance Theory), a Hybrid Model that employs techniques like intelligent operators, clonal selection principle, local search, memory antibodies and ART clusterization, in order to increase the performance of the algorithm. The approach uses a mechanism similar to the ART 1 network for storing a population of memory antibodies that will be responsible for the acquired knowledge of the algorithm. This characteristic allows the algorithm a self-organization of the antibodies in accordance with the complexity of the database.

  • [5] Hruschka E. R. Jr., dos Santos E. B., de O. Galvao S. D. C. (Brazil): An optimized evolutionary conditional independence Bayesian Classifier induction process, 555-572.

    Bayesian Networks (BNs) are graphical models which represent multivariate joint probability distributions which have been used successfully in several studies in many application areas. BN learning algorithms can be remarkably effective in many problems. The search space for a BN induction, however, has an exponential dimension. Therefore, finding the BN structure that better represents the dependencies among the variables is known to be a NP problem. This work proposes and discusses a hybrid Bayes/Genetic collaboration (VOGAC-MarkovPC) designed to induce Conditional Independence Bayesian Classifiers from data. The main contribution is the use of MarkovPC algorithm in order to reduce the computational complexity of a Genetic Algorithm (GA) designed to explore the Variable Orderings (VOs) in order to optimize the induced classifiers. Experiments performed in a number of datasets revealed that VOGAC-MarkovPC required less than 25% of the time demanded by VOGAC-PC on average. In addition, when concerning the classification accuracy, VOGAC-MarkovPC performed as well as VOGAC-PC did.

  • [6] Roselina Sallehuddin, Siti Mariyam Hj. Shamsuddin, Siti Zaiton Mohd. Hashim, Ajith Abraham (Malaysia): Forecasting time series data using hybrid grey relational artificial neural network and auto regressive integrated moving, 573-605.

    In business, industry and government agencies, anticipating future behavior that involves many critical variables for nation wealth creation is vitally important, thus the necessity to make precise decision by the policy makers is really essential. Consequently, an accurate and reliable forecast system is needed to compose such predictions. Accordingly, the aim of this research is to develop a new hybrid model by combining a linear and nonlinear model for forecasting time series data. The proposed model (GRANN_ARIMA) integrates nonlinear Grey Relational Artificial Neural Network (GRANN) and linear ARIMA model, combining new features such as multivariate time series data as well as grey relational analysis to select the appropriate inputs and hybridization succession. To validate the performance of the proposed model, small and large scale data sets are used. The forecasting performance was compared with several models, and these include: individual models (ARIMA, Multiple Regression, Grey Relational Artificial Neural Network), several hybrid models (MARMA, MR_ANN, ARIMA\_ANN), and Artificial Neural Network (ANN) trained using Levenberg Marquardt algorithm. The experiments have shown that the proposed model has outperformed other models with 99.5% forecasting accuracy for small-scale data and 99.84% for large-scale data. The empirical results obtained have proved that the GRANN_ARIMA model can provide a better alternative for time series forecasting due to its promising performance and capability in handling time series data for both small and large scale data.

  • [7] Arijit Biswas, Sambarta Dasgupta, Swagatam Das, Ajith Abraham (India, Norway): A synergy of differential evolution and bacterial foraging optimization for global optimization, 607-626.

    The social foraging behavior of Escherichia coli bacteria has recently been studied by several researchers to develop a new algorithm for distributed optimization control. The Bacterial Foraging Optimization Algorithm (BFOA), as it is called now, has many features analogous to classical Evolutionary Algorithms (EA). Passino [1] pointed out that the foraging algorithms can be integrated in the framework of evolutionary algorithms. In this way BFOA can be used to model some key survival activities of the population, which is evolving. This article proposes a hybridization of BFOA with another very popular optimization technique of current interest called Differential Evolution (DE). The computational chemotaxis of BFOA, which may also be viewed as a stochastic gradient search, has been coupled with DE type mutation and crossing over of the optimization agents. This leads to the new hybrid algorithm, which has been shown to overcome the problems of slow and premature convergence of both the classical DE and BFOA over several benchmark functions as well as real world optimization problems.

  • [8] Köppen M., Yoshida K. (Japan): Many-objective training of a multi-layer perceptron, 627-637.

    In this paper, a many-objective training scheme for a multi-layer feed-forward neural network is studied. In this scheme, each training data set, or the average over sub-sets of the training data, provides a single objective. A recently proposed group of evolutionary many-objective optimization algorithms based on the NSGA-II algorithm have been examined with respect to the handling of such problem cases. A modified NSGA-II algorithm, using the norm of an individual as a secondary ranking assignment method, appeared to give the best results, even for a large number of objectives (up to 50 in this study). However, there was no notable increase in performance against the standard backpropagation algorithm, and a remarkable drop in performance for higher-dimensional feature spaces (dimension 30 in this study).

  • [9] Fatos Xhafa (Spain): A Hyper-heuristic for adaptive scheduling in Computational Grids, 639-656.

    In this paper we present the design and implementation of an hyper-heuristic for efficiently scheduling independent jobs in Computational Grids. An efficient scheduling of jobs to Grid resources depends on many parameters, among others, the characteristics of the resources and jobs (such as computing capacity, consistency of computing, workload, etc.). Moreover, these characteristics change over time due to the dynamic nature of Grid environment, therefore the planning of jobs to resources should be adaptively done. Existing ad hoc scheduling methods (batch and immediate mode) have shown their efficacy for certain types of resource and job characteristics. However, as stand alone methods, they are not able to produce the best planning of jobs to resources for different types of Grid resources and job characteristics.

    In this work we have designed and implemented a hyper-heuristic that uses a set of ad hoc (immediate and batch mode) scheduling methods to provide the scheduling of jobs to Grid resources according to the Grid and job characteristics. The hyper-heuristic is a high level algorithm, which examines the state and characteristics of the Grid system (jobs and resources), and selects and applies the ad hoc method that yields the best planning of jobs. The resulting hyper-heuristic based scheduler can be thus used to develop network-aware applications that need efficient planning of jobs to resources.

    The hyper-heuristic has been tested and evaluated in a dynamic setting through a prototype of a Grid simulator. The experimental evaluation showed the usefulness of the hyper-heuristic for planning of jobs to resources as compared to planning without knowledge of the resource and job characteristics.

  • [10] Lorena A. C., Carvalho A. C. P. L. F. (Brazil): Design of directed acyclic graph multiclass structures, 657-673.

    One of the approaches adopted to generate multiclass classifiers from binary predictors is to decompose the multiclass problem into multiple binary subproblems. Among the existing decomposition approaches, one may cite the use of Directed Acyclic Graphs (DAG) to combine pairwise classifiers. This work presents a study on the influence of the DAG structure in the performance obtained in multiclass problems when Support Vector Machines are used in the induction of the binary predictors.

  • [11] Snášel V.,  Platoš J., Krömer P., Húsek D., Frolov A. (Czech Republic): On the road to genetic Boolean matrix factorization, 675-688.

    Matrix factorization or factor analysis is an important task helpful in the analysis of high dimensional real world data. There are several well known methods and algorithms for factorization of real data but many application areas including information retrieval, pattern recognition and data mining require processing of binary rather than real data. Unfortunately, the methods used for real matrix factorization fail in the latter case. In this paper we introduce background and initial version of Genetic Algorithm for binary matrix factorization.


  • [1] Beghdad R. (Algeria): Training a small KDD subset to detect and classify attacks, 415-429.

    This paper presents a neural network (NN) approach to detect intrusions. Previous works used many KDD records to train NNs for detecting intrusions. That is why; our objective here is to show that in case of the KDD data sets, we can obtain good results by training some NNs with a small data subset. To prove that, this study compares the attacks detection and classification by using two training sets: a set of only 260 records and a set of 65536 records. The testing set is composed of 65536 records randomly chosen from the KDD testing set. Our study focused on two classification types of records: a single class (normal or attack), and a multi class where the category of the attack is detected by the NN. Four different types of NNs were tested: Multi-Layer Perceptron (MLP), Modular, Jordan/Elman and Principal Component Analysis (PCA) NN. Two NN structures were used: the first one contains only one hidden layer and the second contains ten hidden layers. Our simulations show that the small data subset (260 records) can be trained to detect and classify attacks more efficiently than the second data subset.

  • [2] Borji A., Hamidi M. (Iran): Optical character recognition motivated by Primate Visual System, 413-445.

    A visual nervous system inspired approach to optical character recognition is proposed in this paper with the hope to touch human performance in a limited extent. Particularly, the application of features motivated by the hierarchical structure of the visual ventral stream for recognition of both English and Persian handwritten digits is investigated. Features are derived by combining position and scale invariant edge detectors in a hierarchy over neighboring positions and multiple orientations. The extracted features are then used to train and test a classifier. We examine three types of classifiers: ANN, SVM and kNN to show that features are not dependent on a specific classifier which is in support of these features. The evaluation of the proposed method over standard Persian and English handwritten digit datasets shows high recognition rates of 99.63% and 98.9%, respectively. A stability analysis is also performed to demonstrate the robustness of this method to orientation, scale, and translation distortions.

  • [3] Coufal D., Vydra J., Selicharová I. (Czech Republic): GUHA analysis of proteomic oncological data, 447-456.

    The paper presents results of GUHA analysis of proteomic data. The data are related to an oncological study on breast cancer and are given by 2D electrophoresis gels carrying expression intensity of proteins in cancer cells. The gels have been classified by a physician according to the clinical course of the tumor disease. A research task is aimed on search for significant relations between protein spot intensities and respective clinical presentation. The task was solved by the GUHA method of data mining.

  • [4] Nasiri-Avanaki M.-R., Ebrahimpour R. (Iran): In-service video quality measurements in optical fiber links based on neural network, 457-468.

    Video quality assessment plays a key role in evaluating and optimizing video systems. In this paper, objective image quality metric, appropriate for the fiber optic image transmission line without accessing to the original images, is evaluated by conventional neural networks. It allows an efficient continuous time scoring of the video stream efficiently by a mark on a scale of zero to five. The image database in this research has been collected in the Semnan University. The certainty of the trained network is above 81 percent. Simulation results show that the proposed method is highly correlated with experimental data collected through the subjective experiments.

  • [5] Svítek M. (Czech Republic): Wave probabilistic models, 469-481.

    The paper presents the basic theory of wave probabilistic models together with their features. By introduction of the complementarity's principle between x-representation and k-representation the probability theory is completed for "structural" parameter which carries information about the changes of time series or random processes. The next feature of wave probabilistic models is the quantization principle or definition of probabilistic inclusion-exclusion rules.

  • [6] Zhong Shisheng, Li Yang, Ding Gang, Lin Lin (China): Continuous wavelet process neural network and its application, 483-495.

    In this paper, a continuous wavelet process neural network (CWPNN) model is proposed based on the wavelet theory and process neural network model. The network offers good compromise between robust implementations resulting from the redundancy characteristic of non-orthogonal wavelets, and efficient functional representations that build on the time-frequency localization property of wavelets. Moreover, the network can deal with continuous input signals directly. The corresponding learning algorithm is given and the network is used to solve the problems of aeroengine condition monitoring. The simulation test results indicate that the CWPNN has a faster convergence speed and higher accuracy than the same scale process neural network (PNN) and BP neural network. This provided an effective way for the problems of aeroengine condition monitoring.

  • [7] Zitar R. A., Hanandeh E., Shehabat E. (Jordan): Genetic algorithm with neighbor solution approach for Traveling Salesman Problem, 497-504.

    The Traveling Salesman Problem (TSP) is an NP-Complete problem. Many techniques were developed to solve such problem, including Genetic Algorithms (GA's). The goal of this research is to enhance the performance of the Genetic Algorithm (GA) in solving the TSP. We achieve this goal by developing a local search algorithm called Search for Nearest Solution Algorithm (SNSA). This algorithm produces better solutions in acceptable periods of time. A new Crossover operator is proposed and used in the SNSA. Results for benchmark cases of the TSP show that the algorithm can find known optimum solutions. Comparisons between the proposed algorithm and other GA based methods with known crossover techniques show that SNSA has good quality solutions.


  • [1] Editorial, 269-270.
  • [2] Hussain B. A., Al-Dabbagh R. D. (Iraq/Baghdad): A canonical genetic algorithm for likelihood estimator of first order moving average model parameter, 271-285.

    The increasing availability of computing power in the past two decades has been used to develop new techniques for optimizing the solution of estimation problem. Today's computational capacity and the widespread availability of computers have enabled the development of a new generation of intelligent computing techniques, such as the algorithm of our interest. This paper presents a new member of the class of stochastic search algorithms (known as Canonical Genetic Algorithm "CGA") for optimizing the maximum likelihood function ln (L(θ, σa2 )) of the first order moving average MA(1) model. The presented strategy is composed of three main steps: recombination, mutation, and selection. The experimental design is based on simulating the CGA with different values of (θ), and sample size n. The results are compared with those of moment method. Based on MSE value obtained from both methods, one can conclude that CGA can give estimators (\hat \theta) for MA(1) parameter which are good and more reliable than those estimators obtained by moment method.

  • [3] Temel T., Karlik B. (Turkey): An improved odor recognition system using learning vector quantization with a new discriminant analysis, 287-294.

    A new pre-processing algorithm for improved discrimination of odor samples is proposed. The pre-processed odor sample outputs from two sensors are input using a learning-vector quantization (LVQ) classifier as a means of odor recognition to be employed within electronic nose applications. The proposed algorithm brings out highly scattered classes while minimizing the within-class scatter of the samples given an odor class. LVQ is observed to operate robustly and reliably in terms of variation of parameters of interest, mainly a learning parameter. Due to the increased performance along with computational simplicity and robustness, the scheme is suitable to sample-by-sample identification of olfactory sensory data and can be easily adapted to hierarchical processing with other sensory data in real-time.

  • [4] Owais S. S. J., Snášel V., Krömer P. (Jordan, Czech Republic): Grow up precision recall relationship curve in IR system using GP and fuzzy optimization in optimizing the user query, 295-309.

    An information retrieval (IR) system (IRs) (search engine) is said to be efficient, to the degree that always evaluates each object in the information base (database, document base, web,...) like the expert. The ability of IRs's is to retrieve mostly all relevant objects (measured by the recall), and only the (most) relevant objects (measured by the precision) from the collection queried.

    Recall and precision measures provide the classical measure of the retrieval efficiency. They measure the degree to which the query answer (the set of documents that retrieved by IRs as response to the user query). Where, the query answer is the set of relevant documents in the information based queried.

    Retrieving most relevant documents to the user query in IRs was one of the most important methods of World Wide Web (WWW) search engines used in the world now. So the searchers aim to use genetic programming (GP) and fuzzy optimization to optimize the user search query in the Boolean IRs model and in the fuzzy IRs model; and to use more Boolean operators (AND, OR, XOR, OF, and NOT) instead of using the standard operators (AND, OR, and NOT), and to use weights for terms and for Boolean operators. Weights are used to give the users more relaxation in defining how much the importance of the terms and of the Boolean operators is. The terms and the Boolean operators' weights are used in fuzzy IRs model. In addition, it investigates extensions of the classical measurement of effectiveness in IRs, precision; recall and harmonic mean.

    The researchers use harmonic mean measure as an objective function which uses both measures precision and recall at once for evaluating the results of the two IRs models to grow up the precision-recall relationship curve.

  • [5]  Sleit A., Al-Mbaideen W., Alzabin N., Dawood H., Alqarute K. (Jordan): Efficient query processing over mirror servers using genetic algorithms, 311-320.

    In a mirror server environment, clients request services from servers. Therefore, the system must have an intelligent algorithm to select the most suitable server to fulfill a coming request. Choosing such a server for a particular client may be very difficult. Evolutionary techniques can be utilized to determine the server best suited to a particular client request based on parameters such as processing and reply times. Usage of genetic algorithms in server selection is researched in this paper taking into consideration various probabilities for mutation and crossover.

  • [6] Snášel V., Krömer P., Owais S. S. J., Nyongesa H. O., Maleki-Dizaji S. (Czech Republic, Jordan, United Kingdom): Evolutionary improving World Wide Web queries, 321-334.

    As the volume and variety of information sources, especially on the World Wide Web (WWW), continue to grow, the requirements imposed on search applications are steadily increasing. The amount of available data is growing and so do the user demands. Search application should provide the users with accurate, sensible responses to their requests. It is difficult to provide information that accurately matches user information needs. Search effectiveness can be seen as the accuracy of matching user information needs against the retrieved information. There are problems emerging: users often do not present search queries in the form that optimally represents their information need, the measure of a document's relevance is often highly subjective between different users, and information sources might contain heterogeneous documents, in multiple formats and the representation of documents is not unified. This contribution presents a proposal to improve web search effectiveness via evolutionary optimization of the Boolean and vector search queries based on individual user models.

  • [7] Barazane L., Khwaldeh A., Jumah M., El-Qawasmeh E. (Algeria, Jordan): Neuro-fuzzy sliding mode control applied to the field-oriented control of an induction motor, 335-349.

    The aim of this paper consists in using one of the emergent techniques which proves its capability of improving performances of several systems, called "neuro-fuzzy", in order to reduce the chattering phenomenon and also to perform the control obtained with fuzzy sliding mode control. In fact, after determining the decoupled model of the motor, a set of simple surfaces and associated a smooth control function with a threshold have been synthesized. However, the magnitude of this control function depends closely on the upper bound of uncertainties, which include parameter variations and external disturbances, and this generates chattering. Usually, the upper bound of uncertainties is difficult to be known before motor operation, so a fuzzy sliding mode controller is investigated to solve this difficulty and in which a simple fuzzy inference mechanism is used to reduce the chattering phenomenon by simple adjustments. In order to optimize the control performances and ensure a significant reduction of chattering compare with ones obtained in the previous fuzzy sliding mode, we propose in this paper to use adaptive predictive neural approach to regulate the speed of the motor. The neural control algorithm is provided with the predicted system output which is the speed variable via a recursive on line identification of the overall system which is based on a static feed forward linear network with one hidden layer. The predicted data are passed to a numerical optimization algorithm which attempts to minimize a quadratic performance criterion to compute the suitable control signal.

  • [8] Shanmugam B., Idris N. B. (Malaysia): Improved hybrid intelligent intrusion detection system using AI technique, 351-362.

    Intrusion detection systems are increasingly a key part of systems defense. Various approaches to intrusion detection are currently being used, but they are relatively ineffective. Artificial Intelligence plays a driving role in security services. This paper proposes a dynamic model of intelligent intrusion detection system, based on a specific AI approach for intrusion detection. The techniques that are being investigated include fuzzy logic with network profiling, which uses simple data mining techniques to process the network data. The proposed hybrid system combines anomaly and misuse detection. Simple fuzzy rules allow us to construct if-then rules that reflect common ways of describing security attacks. We use DARPA dataset for training and benchmarking.

  • [9] Shamsuddin S. M., Darus M., Saman F. I. (Malaysia): Three term backpropagation algorithm for classification problem, 363-376.

    Standard Backpropagation Algorithm (BP) usually utilizes two term parameters; Learning Rate α and Momentum Factor β. Despite the general success of this algorithm, there are several drawbacks such as existence of local minima, slow rates of convergence and modification of algorithm requires complex computations. In this study, further analysis of proportional factor γ for 3-Term BP is investigated on various scales of datasets; small, medium and large. Experiments are conducted using three UCI dataset; Balloon, Iris and Cancer. The results show that the 3-Term BP outperforms standard BP for small scale data, but does not work well for medium and large scale dataset.

  • [10] Al-Jowher W. A., Al-Ramahi N. N., Alfaouri M. (Amman-Jordan): Image identification and labeling using hybrid transformation and neural network, 377-395.

     Face recognition has a wide range of applications such as personal identification and authentication, criminal identification, security and surveillance, image and film processing, and human-computer interaction. Although there exist many methods, this paper proposes recent face recognition using a dynamic programming algorithm for image recognition and classification. A method based on a new mapping network called wavelet-network, namely Wavenet transform (WN). WN was employed to make approximation to the images before passing through the discrete wavelet transform decomposition to extract the image descriptive features. These features are used in the proposed image identification algorithm for enhancing the accuracy of recognition at pixel level and to minimize the additive cost function.

    The proposed hybrid transform is based on the combination of the Wavenet (WN) and the Inverse Discrete Wavelet Transform (IDWT) followed by a Neural Network (NN) to be considered as feature extractor for the given image. In this paper the neural network (NN) classifier is combined with the wavelet transform. A reference set of 100 images are used and collected from different data images. This method gave an excellent and successful identification rate of 99%. Gaussian noise was added for further testing, the proposed algorithm for the same collected images and identification rate of 95% was achieved with level of up to 0.10.

    The algorithm was implemented using MATLAB programming languages version 7.

  • [11] Peyravi F., Pashaei K., Taghiyareh F. (Iran): A multi agent community of practice, 397-413.

    Modern organizations tend to constitute of communities of practice to cover the side effect of standardization and centralization of knowledge. The distributed nature of knowledge in groups, teams and other departments of organization and complexity of this tacit knowledge lead us to use community of practice as an environment to share knowledge. In this paper we propose an agent mediated community of a practice system using MAS-CommonKADS methodology. We support the principle of autonomy since every single agent, even those in the same community, needs its own autonomy in order to model an organization and its individuals correctly, using this approach, the natural model for an agent based on knowledge sharing system has been resulted. We presented all models of MAS-CommonKADS methodology required for developing the multi-agent system. We found MAS-CommonKADS useful to design Knowledge Management applications. Because of detailed description of agents, a resulted design model could be simply implemented. We modeled our system using Rebeca and verified it to show that by use of our system, knowledge sharing can be satisfied.


  • [1] Alata M., Moaqet H. (Jordan): Adaptive neuro-fuzzy inference system with second order Sugeno consequents, 171-187.

    Adaptive Neuro-Fuzzy Inference System (ANFIS) with first order Sugeno consequent is used widely in modeling applications. Though it has the advantage of giving good modeling results in many cases, it is not capable of modeling highly non-linear systems with high accuracy. In this paper, an efficient way for using ANFIS with Sugeno second order consequents is presented. Better approximation capability of Sugeno second order consequents compared to lower order Sugeno consequents is shown. Subtractive clustering is used to determine the number and type of membership functions. A hybrid-learning algorithm that combines the gradient descent method and the least squares estimate is then used to update the parameters of the proposed Second Order Sugeno-ANFIS (SOS-ANFIS). Simulation of the proposed SOS-ANFIS for two examples shows better results than that of lower order Sugeno consequents. The proposed SOS-ANFIS shows better initial error, better convergence, quicker convergence and much better final error value.

  • [2] Holeňa M. (Czech Republic): McNaughton theorem of fuzzy logic from a data-mining point of view, 189-212.

    The paper recalls the McNaughton theorem of fuzzy logic and the algorithms underlying its constructive proofs. It then shows how those algorithms can be combined with the algorithm underlying recent extension of the theorem to piecewise-linear functions with rational coefficients, and points out potential importance of the resulting combined algorithm for data mining. That result is immediately weakened through a complexity analysis of the algorithm that reveals that its worst-case complexity is doubly-exponential.

  • [3] Kilic E., Dundar P. (Turkey): The edge-accessibility number via graph, operations & an algorithm, 213-223.

    In communications network design, network's stability is a very important concept. A network has to be constructed as possible as stable since the stability of a network shows its resistance to vulnerability. Many science and engineering problems can be represented by a network, generalization of which is a graph. Examples of problems that can be represented by a graph include: cyclic sequential circuit, organic molecule structures, mechanical structures, etc. So, a graph can be considered as a model of a communication network. Then, the notions of the graph theory can be used for the stability of a network. In the graph theory, deterministic measures of the stability are used for some parameters of graphs as connectivity, covering number, independence number and dominating number. Then, the stability of a network is defined with deterministic calculation. Today, these parameters take into consideration the neighborhood notion. Now, we consider an edge-accessibility number of a graph. Edge-accessibility is a notion which uses the neighborhood of edges (links). In this paper; we search the edge-accessibility number of a graph. We also give some theorems about the edge-accessibility using the graph operations and design an algorithm which found it with Time Complexity O(n3).

  • [4] Senouci M., Liazid A., Beghdadi H. A., Benhamamouch D. (Algeria): A segmentation method to handwritten word recognition, 225-236.

    This paper presents a segmentation technique to handwritten word recognition. This technique implements an algorithm based on an analytical approach. It uses a letter sweeping procedure with a step equal to the Euclidean distance between an established reference index and the entity (the alphabet letter). Then a dissociation of this entity is achieved when this distance will reach a rate of 80%. Our experience about this segmentation technique gives a rate of 81.05% of recognition. A neural multi-layer perceptron classifier confirms the extracted segment. This procedure is successively repeated from the beginning until the end of the word. A concatenation technique is finally used to the word reconstitution.

  • [5] Svítek M. (Czech Republic): Complementary variables and its application in statistics, 237-253.

    The paper presents the basic theory of complementary statistics and its application in the area of applied probabilistic modeling. By introduction of the complementarity's principle between x-representation (random time series, random process) and p-representation or k-representation (rate of change/velocity of random time series and processes) the probability theory is completed for the "structural" parameter which carries information about the changes of studied time series or the random process. At the end, the basic application of probabilistic modeling is introduced and the presented principle is illustrated on the set of numerical examples with different probability density functions.

  • [6] Übeyli E. D. (Turkey): Wavelet/probabilistic neural networks for ECG beats classification, 255-267.

    A new approach based on the implementation of probabilistic neural network (PNN) is presented for classification of electrocardiogram (ECG) beats. Four types of ECG beats (normal beat, congestive heart failure beat, ventricular tachyarrhythmia beat, atrial fibrillation beat) obtained from the Physiobank database were analyzed. The ECG signals were decomposed into time-frequency representations using discrete wavelet transform (DWT) and wavelet coefficients were calculated to represent the signals. The aim of the study is classification of the ECG beats by the combination of wavelet coefficients and PNN. The purpose is to determine an optimum classification scheme for this problem and also to infer clues about the extracted features. The present research demonstrated that the wavelet coefficients are the features which well represent the ECG signals and the PNN trained on these features achieved high classification accuracies.


  • [1] Beghdad R. (Algeria): Training all the KDD data set to classify and detect attacks, 81-91.

    The purpose of this study is to analyze the performances of some neural networks (NNs) when all the KDD data set is used to train them, in order to classify and detect attacks. Five different types of NNs were tested: Multi-Layer Perceptron (MLP), Self Organization Feature Map (SOFM), Radial Basis Function/Generalized Regression/Probabilistic (RBF/GR/P), Jordan/Elman, and Recurrent NNs. The experiment study is done on the Knowledge Discovery and Data mining (KDD) data sets. We consider two levels of attack granularities depending on whether dealing with four main categories, or only focusing on the normal/attack connection types. Our simulations show that our results are competitive with some other artificial intelligence or data mining intrusion detection systems.

  • [2] Dunis C. L., Nathani A.(United Kingdom): Quantitative trading of gold and silver using nonlinear models, 93-111.

    The main aim of this paper is to forecast gold and silver daily returns with advanced regression analysis using various linear and non-linear models.

    ARMA models are used as a linear benchmark for comparison purposes with established non-linear models such as Nearest Neighbours and MultiLayer Perceptron (MLP), and Higher Order Neural Networks (HONN) whose application to financial markets is quite new. All models are assessed using statistical criteria such as correct directional change as well as financial criteria such as risk adjusted return. The main aim is to find which of these models generate the best returns and if nonlinear models can be used for generating excess returns in the precious metals market. This is achieved by implementing a trading simulation where the forecast is translated into a trading signal. Profit statistics are calculated taking into account transaction costs.

    It is concluded that, for the January 2000-May 2006 period under review, nonlinear models like MLPs and HONNs did outperform the linear ARMA models. In the end, the performance of both MLP and HONN models showed the presence of nonlinearities in the gold and silver prices as it was found that nonlinear models can be effectively used for generating excess returns in these markets.

  • [3] Venkatalakshmi K., Shalinie S. M. (India): Multispectral image classification using modified k-means algorithm, 113-120.

    Clustering is used to organize data for efficient retrieval. A popular technique for clustering is based on k-Means such that the data is partitioned into k clusters. In k-Means clustering a set of n data points in d-dimensional space Rd, an integer k is given and the problem is to determine a set of k-points in Rd called centers, to minimize the mean squared distance from each point to its nearest center. In this method, the number of clusters is predefined and the technique is highly dependent on the initial identification of elements that represent the clusters well. A large area of research in clustering has focused on improving the clustering process such that the clusters are not dependent on the initial identification of cluster representation. In this paper, a modified technique, which grows the clusters without the need to specify the initial cluster representation, has been proposed. Initially a local search single swap heuristic can identify the number of clusters and its centers in the interpolated (bicubic) multispectral image. Then the regular k-Means clustering is implemented using the results of the previous process for the true image data set. The technique achieves an impressive speed up of the clustering process even when the number of clusters is not specified initially and the classification accuracy is improved within a fewer number of iterations.

  • [4] Jiuwen Zhang, Kun Zhan, Yide Ma (China): Rotation and scale invariant antinoise PCNN features for content-based image retrieval, 121-132.

    Content-based image retrieval (CBIR) of images, especially those with different orientation, scale changes and noise affects, are a challenging and important problem in the image analysis. This paper proposes an effective scheme for rotation and scale invariant antinoise retrieval using pulse-coupled neural network (PCNN) features. The PCNN generates series of pulse images, which are binary and represent different features of the original image. The series of pulse images can be then calculated to an entropy sequence called the feature of the image. The experimental results show that the retrieval scheme is effective in extracting rotation and scale invariant features and it also performs better robustness to noise.

  • [5] Babinec Š., Pospíchal J. (Czech Republic): Optimization of echo state neural networks for electrical load forecasting, 133-152.

    The predictive performance of Echo State neural networks were optimized for electrical load forecasting and compared to the results achieved by competitors in the worldwide Eunite Competition #1. The test data used were the actual results of the competition, attached to a specific region. A regular adaptation of an \textit{Echo State }neural network was optimized by adapting the weights of the dynamic reservoir through Anti-Hebbian learning, and the weights from input and output neurons to the hidden neurons were optimized using the Metropolis algorithm. The results achieved with such an optimized Echo State neural network would gain a strong second place within the Eunite competition.

  • [6] Guney K., Onay M. (Turkey): Bees algorithm for null synthesizing of linear antenna arrays by controlling only the element positions, 153-169.

    This paper presents bees algorithm (BA) for null steering of linear antenna arrays by controlling only the element positions. The BA is an optimization algorithm inspired by the natural foraging behavior of honey bees to find the optimal solution. To show the versatility and flexibility of the proposed BA, several examples of Chebyshev array pattern with the imposed single, multiple and broad nulls are given. It is found that the nulling technique based on BA is capable of steering the array nulls precisely to the undesired interference directions. For practical consideration, the sensitivity of the produced patterns due to small variations of the element positions is also examined by rounding the element position values to the second decimal position.


  • [1] Beghdad R. (Algeria): Applying Fisher's Filter to Select KDD Connections' Features and Using Neural Networks to Classify and Detect Attacks, 1-16.

    Most of the neural networks-based intrusion detection systems (IDS) examine all data features to detect intrusion or misuse patterns. Some of the features may be redundant or contribute little (if anything) to the detection process. That is why the purpose of this study is to identify important KDD features which will be used to train a neural network (NN), in order to best classify and detect attacks. Four NNs were studied: Modular, Recurrent, Principal Component Analysis (PCA), and Time-Lag recurrent (TLR) NNs. We investigated the performance of combining the Fisher's filter used as a feature selection technique, with one of the previously cited NNs. Our simulations show that using Fisher's filter improves largely the performance of the four considered NNs in terms of detection rate, attack classification, and computational time.

  • [2] Kontrimas V., Verikas A. (Lithuania): Neural networks based screening of real estate transactions, 17-30.

    Aiming to hide the real money gains and to avoid taxes, fictive prices are sometimes recorded in the real estate transactions. This paper is concerned with artificial neural networks based screening of real estate transactions aiming to categorize them into "clear" and "fictitious" classes. The problem is treated as an outlier detection task. Both unsupervised and supervised approaches to outlier detection are studied here. The soft minimal hyper-sphere support vector machine (SVM) based novelty detector is employed to solve the task without the supervision. In the supervised case, the effectiveness of SVM, multilayer perceptron (MLP), and a committee based classification of the real estate transactions are studied. To give the user a deeper insight into the decisions provided by the models, the real estate transactions are not only categorized into "clear" and "fictitious" classes, but also mapped onto the self organizing map (SOM), where the regions of "clear", "doubtful" and "fictitious" transactions are identified. We demonstrate that the stability of the regions evolved in the SOM during training is rather high. The experimental investigations performed on two real data sets have shown that the categorization accuracy obtained from the supervised approaches is considerably higher than that obtained from the unsupervised one. The obtained accuracy is high enough for the technique to be used in practice.

  • [3] Xiaolin Li, Jinde Cao (China): Exponential Stability of Stochastic Interval Hopfield Neural Networks with Time-Varying Delays, 31-40.

    In this paper, stochastic interval Hopfield neural networks with time-varying delays are investigated. By applying the Razumikhin-type theorem as well as inequality technique, a set of novel sufficient criteria independent of delays are given for the exponential stability of such networks. As a by-product, for the deterministic Hopfield neural networks with time-varying delays, some delay-independent criteria for their global exponential robust stability are also obtained. The proposed results improve and extend them in the earlier literature and are easier to verify. A numerical example and simulation are also given to illustrate the effectiveness of our results.

  • [4] He-Sheng Tang, Song-Tao Xue, Rong Chen, Tadanobu Sato (Japan): HY Filtering Method for Neural Network Training and Pruning, 41-63.

    An efficient training and pruning method based on the HY filtering algorithm is proposed for feedforward neural networks (FNN). A FNN's weight importance measure linking up prediction error sensitivity obtained from the HY filtering training, and then a weight salience based pruning algorithm is derived. Moreover, based on the monotonicity property of the HY filtering Riccati equation and the initial value of the error covariance matrix, performance of the HY filtering training algorithm will also be investigated. The simulation results show that our approach is an effective training and pruning method for neural networks.

  • [5] Dalalah D., Hayajneh M. T. (Jordan): Dynamic Neural Model for Mobile Robot Navigation, 65-80.

    A topologically oriented neural network is very efficient in real-time path planning of a mobile robot in dynamic environments. Using a dynamic recurrent neural network to solve the partial differential equation of a potential field in a discrete manner, the problem of obstacle avoidance and path planning of a moving robot can be efficiently solved. A dimensional network used to represent the topology of the robot's workspace, where each network node represents a state associated with a local workspace point. In this paper, two approaches associated with different boundary conditions are proposed, namely, Dirichlet and Neumann conditions. The first approach relies on a field of attraction distributed around the moving target, acting as a unique local extreme in the local network space. The steepest gradients of the network state variables will aim towards the source of the potential field. The second approach considers two attractive and repulsive potential sources associated with the start and destination points. A dynamic neural mesh is used to model the robot workspace. A simulation package has been built and extensive computer experiments were conducted to demonstrate and validate the reliability of the presented approach.

Thanks to