## Contents of Volume 23 (2013)

1/2013 2/2013 3/2013 4/2013 5/2013### 6/2013

- [1]
**Li Liu, Xiwei Chen, Dashi Luo, Yonggang Lu, Guandong Xu, Ming Liu: HSC:**

A spectral clustering algorithm combined with hierarchical method, 499-521.

**First page Full text DOI: 10.14311/NNW.2013.23.031**

Abstract: Most of the traditional clustering algorithms are poor for clustering more complex structures other than the convex spherical sample space. In the past few years, several spectral clustering algorithms were proposed to cluster arbitrarily shaped data in various real applications. However, spectral clustering relies on the dataset where each cluster is approximately well separated to a certain extent. In the case that the cluster has an obvious inflection point within a non-convex space, the spectral clustering algorithm would mistakenly recognize one cluster to be different clusters. In this paper, we propose a novel spectral clustering algorithm called HSC combined with hierarchical method, which obviates the disadvantage of the spectral clustering by not using the misleading information of the noisy neighboring data points. The simple clustering procedure is applied to eliminate the misleading information, and thus the HSC algorithm could cluster both convex shaped data and arbitrarily shaped data more efficiently and accurately. The experiments on both synthetic data sets and real data sets show that HSC outperforms other popular clustering algorithms. Furthermore, we observed that HSC can also be used for the estimation of the number of clusters.

- [2]
**Chiroma H., Abdulkareem S., Abubakar A., Joda Usman M.:**

Computational Intelligence Techniques with Application to Crude Oil Price Projection: A Literature Survey from 2001- 2012, 523-551.

**First page Full text DOI: 10.14311/NNW.2013.23.032**

Abstract: This paper is an attempt to survey the applications of computational intelligence techniques for predicting crude oil prices over a period of ten years. The purpose of this research is to provide an exhaustive overview of the existing literature which may assist prospective researchers. The reviewed literature covers a spectrum of publications on the proposed model, source of experimental data, period of data collection, year of publication and contributors. The overall trend of the publications in this area of research issued within the last decade is also addressed. The existing body of research has been analyzed and new research directions have been outlined that have been previously ignored. It is expected that researchers across the globe may thus be encouraged to redirect their attention and resources in order to keep on searching for an optimum solution.

- [3]
**Sabri Kaya, Kerim Guney, Celal Yildiz, Mustafa Turkmen:**

ANFIS models for synthesis of open supported coplanar waveguides, 553-569.

**First page Full text DOI: 10.14311/NNW.2013.23.033**

Abstract: Simple and accurate models based on adaptive-network-based fuzzy inference system (ANFIS) to compute the physical dimensions of open supported coplanar waveguides are presented. The ANFIS is a class of adaptive networks which are functionally equivalent to fuzzy inference systems. Four optimization algorithms, hybrid learning, simulated annealing, least-squares, and genetic, are used to determine optimally the design parameters of the ANFIS. When the performances of ANFIS models are compared with each other, the best results are obtained from the ANFIS models trained by the hybrid learning algorithm. The results of ANFIS are compared with the results of the conformal mapping technique, the rigorous spectral-domain hybrid mode analysis, the improved spectral domain approach, the synthesis formulas, a full-wave electromagnetic simulator IE3D, and experimental works realized in this study.

- [4]
**Ming-Fu Hsu, Ping-Feng Pai, Wei-Shih Chung:**

A relevance vector machine with rough set theory model in analyzing the life cycle of new economic firms, 571-586.

**First page Full text DOI: 10.14311/NNW.2013.23.034**

Abstract: The subprime mortgage crisis and subsequent financial tsunami have raised considerable concerns about financial risk management and evaluation. This is nowhere more apparent than in new economic firms (NEFs) with large economic targets and heavy R&D expenses, such as firms in the electronics industries. With its potential for extreme growth and superior profitability, the electronic industries in Taiwan have been in the financial stock market spotlight. Recently, the relevance vector machine (RVM) was reported to have considerably less computation complexity than support vector machines (SVM) models, since it uses fewer kernel functions. Another emerging technique is rough set theory (RST), which derives rules from data. Based on the corporation life cycle theory (CLC), this study developed a relevance vector machine with rough set theory (RVMRS) to predict the status of a corporation in the decline stage. To demonstrate the performance of the designed RVMRS model, the study used electronic industries data from the Taiwan Economic Journal data bank, Taiwan Security Exchange, and Securities and Futures Institute in Taiwan. Experimental results revealed that the presented RVMRS model can predict the decline stage in a firms life cycle with satisfactory accuracy, and generate rules for investors, managers, bankers and regulators that enable them to make suitable judgments. In addition, this study proved that the transparency and information disclosure index (TDI) is crucial to predicting the financial decline of corporations.

- [5]
**Kisi O., Aytek A.:**

Explicit neural network in suspended sediment load estimation, 587-607.

**First page Full text DOI: 10.14311/NNW.2013.23.035**

Abstract: Correct estimation of sediment volume carried by a river is very important for many water resources projects. Traditionally, artificial neural networks (ANNs) are used as black-box models without understanding what happens inside the box. The question is that, how anyone who may be unfamiliar with ANNs can apply this kind of models in any other study, while the model has not been formulated. This paper proposes an explicit neural network (ENN) formulation which is simple and can be used, by anyone who is even not familiar with ANNs, for modeling daily suspended sediment-discharge relationship. The daily streamflow and suspended sediment data from two stations on Tongue River in Montana are used as case studies. Two different sediment rating curves (SRC), multi-linear regression (MLR) and nonlinear regression (NLR) are also applied to the same data. The ENN estimates are compared with those of the SRC, MLR and NLR models. The root mean square errors (RMSE), mean absolute errors (MAE), correlation coefficient (R) and model efficiency (E) statistics are used to evaluate the performance of the models. The comparison results reveal that the suggested model performs better than the conventional SRC, MLR and NLR.

- [6]
**Schejbalová Z., Mičunek T., Schmidt D.:**

Biomechanical analysis of the dummy responses in case of child pedestrian/cyclist collision with passenger car, 609-622.

**First page Full text DOI: 10.14311/NNW.2013.23.036**

Abstract: The safety of pedestrians and cyclists in traffic is justified especially in terms of prevention. This paper deals with the biomechanical analysis of load exerted on the child pedestrian and cyclist. In the case of cyclists, the impact configurations were chosen with respect to the statistical outputs (sudden enter the road or the case of non-giving way; the car front vs. the left side of the cyclists). Two tests were performed in the same configuration and nominal collision speed, the first one with a bicycle helmet and the second one without the helmet. The initial position of pedestrian was chosen with respect to the dummy degrees of freedom. Using the accelerometers in the head, chest, pelvis and knee of the dummy acceleration fields were detected, which are the child pedestrian and cyclist exposed during the primary and secondary collision. In addition, prediction diagnostics method implementation was discussed such as one possible solution of vulnerable road users harm reduction. In conclusion, the results are interpreted by values of biomechanical load and severity of potential injuries including kinematic and dynamic comparison.

- [7]
**Contents volume 23 (2013), 623-625.** - [8]
**Author's index volume 23 (2013), 627-629.**

### 5/2013

- [1]
**Garcia-Laencina P. J., Roca-Gonzalez J. L., Bueno-Crespo A., J-L. Sancho-Gomez:**

Exploiting diversity of neural network ensembles based on extreme learning machine, 395-409

**First page Full text DOI: 10.14311/NNW.2013.23.024**

Extreme learning machine (ELM) is an emergent method for training single hidden layer feedforward neural networks (SLFNs) with extremely fast training speed, easy implementation and good generalization performance. This work presents effective ensemble procedures for combining ELMs by exploiting diversity. A large number of ELMs are initially trained in three different scenarios: the original feature input space, the obtained feature subset by forward selection and different random subsets of features. The best combination of ELMs is constructed according to an exact ranking of the trained models and the useless networks are discarded. The experimental results on several regression problems show that robust ensemble approaches that exploit diversity can effectively improve the performance compared with the standard ELM algorithm and other recent ELM extensions.

- [2]
**Jagric T., Zunko M.:**

Neural network world: Optimized spiral spherical SOM (OSS-SOM), 411-426

**First page Full text DOI: 10.14311/NNW.2013.23.025**

The border effect is one of the problems, which can appear in the application of self-organizing maps (SOM). Different solutions were presented in the literature, but each of them has its drawbacks. In this paper we present a new method for overcoming the border effect optimized spiral spherical SOM. We also show that standard measure of irregularity is not appropriate and present a modified version Gaussian measure of irregularity. Our simulations suggest that the new variant of SOM achieves extremely low values of irregularity in comparison to other methods. At the end of the paper we present a software solution for the proposed method.

- [3]
**Tylová L., Kukal J., Vyšata O.:**

Spectral analysis of predictive error in Alzheimer's disease diagnostics, 427-434

**First page Full text DOI: 10.14311/NNW.2013.23.026**

The new method is based on non-linear one-step predictor, which is designed as MLP neural network. It is a kind of low-pass non-linear filter. The difference between raw EEG and the ANN output is then a subject of band spectral analysis. The differences in this power spectrum between Alzheimers diseased and control patient group are statistically significant.

- [4]
**Oladipupo O. O., Uwadia O. C., Ayo K. C., Abraham A., Snášel V.:**

A fuzzy-mining approach for solving rule based expert system unwieldiness in medical domain, 435-450

**First page Full text DOI: 10.14311/NNW.2013.23.027**

Over the years, one of the challenges of a rule based expert system is the possibility of evolving a compact and consistent knowledge-base with a fewer numbers of rules that are relevant to the application domain, in order to enhance the comprehensibility of the expert system. In this paper, the hybrid of fuzzy rule mining interestingness measures and fuzzy expert system is exploited as a means of solving the problem of unwieldiness and maintenance complication in the rule based expert system. This negatively increases the knowledge-base space complexity and reduces rule access rate which impedes system response time. To validate this concept, the Coronary Heart Disease risk ratio determination is used as the case study. Results of fuzzy expert system with a fewer numbers of rules and fuzzy expert system with a large numbers of rules are presented for comparison. Moreover, the effect of fuzzy linguistic variable risk ratio is investigated. This makes the expert system recommendation close to human perception.

- [5]
**Jiřina M., Jiřina M. jr.:**

GMDH method with genetic selection algorithm and cloning, 451-464

**First page Full text DOI: 10.14311/NNW.2013.23.028**

The GMDH MIA algorithm uses linear regression for adaptation. We show that Gauss-Markov conditions are not met here and thus estimations of network parameters are biased. To eliminate this we propose to use cloning of neuron parameters in the GMDH network with genetic selection and cloning (GMC GMDH) that can outperform other powerful methods. It is demonstrated on tasks from the Machine Learning Repository.

- [6]
**Garlík B., Křivan M.:**

The use of evolutionary algorithms to optimize intelligent buildings electricity supply, 465-481

**First page Full text DOI: 10.14311/NNW.2013.23.029**

This paper formulates a unit commitment optimization problem for renewable energy sources distributed in a micro-grid formed by a complex of intelligent buildings of both office and residential characters, including a wide range of amenities. We present a general description of the solution of this task using the simulated annealing heuristic optimization technique. The experiment was processed in the specialized computer program. For comparison, Appendix A of the article describes the Lagrange multipliers optimization method as the conventional alternative to the used heuristic technique. A description of the concept of intelligent buildings is provided in Appendix B.

- [7]
**Bayram B., Acar U., Koca H. K., Narin B., Cavdaroglu G. C., Celik L., Cubuk R.:**

An efficient algorithm for automatic tumor detection in contrast enhanced breast MRI by using artificial neural network (NEUBREA), 483-498

**First page Full text DOI: 10.14311/NNW.2013.23.030**

The advances in image processing technology contribute to the interpretation of medical images and early diagnosis. Moreover various studies can be found in medical journals dedicated to Artificial Neural Networks (ANN). In the presented study, a method was developed to learn and detect benign and malignant tumor types in contrast-enhanced breast magnetic resonance images (MRI). The backpropagation algorithm was taken as the ANN learning algorithm. The algorithm (NEUBREA) was developed in C# programming language by using Fast Artificial Neural Network Library (FANN). Having been diagnosed by radiologists, 7 cases of malignant tumor, 8 cases of benign tumor, and 3 normal cases were used as a training set. The results were tested on 34 cases that had been diagnosed by radiologists. After the comparison of the results, the overall accuracy of algorithm was defined as 92%.

### 4/2013

**Editorial, 285**>

- [1]
**Svítek M.:**

Quantum Subsystems Connections, 287-298.

**First page Full text DOI: 10.14311/NNW.2013.23.018**

Paper presents the results in quantum informatics where two or more quantum subsystems are connected. For modelling the links amongst quantum subsystems the quantum quasi-spin is the most important parameter. We derive a quantum quasi-spin from the condition of logical requirement for the unambiguousness of wave probabilistic function assigned into quantum subsystem. With respect to these results we can define information bosons with integer quasi-spin, information fermions with half-integer quasi-spin and information quarks with third-integer quasi-spin. The methodology can be extended to other variants of quasi-spin.

- [2]
**Štefka D., Holeňa M.:**

Performance of classifcation confdence measures in dynamic classifer systems, 299-320.

**First page Full text DOI: 10.14311/NNW.2013.23.019**

Classifier combining is a popular technique for improving classification quality. Common methods for classifier combining can be further improved by using dynamic classification confidence measures which adapt to the currently classified pattern. However, in the case of dynamic classifier systems, the classification confidence measures need to be studied in a broader context as we show in this paper, the degree of consensus of the whole classifier team plays a key role in the process. We discuss the properties which should hold for a good confidence measure, and we define two methods for predicting the feasibility of a given classification confidence measure to a given classifier team and given data. Experimental results on 6 artificial and 20 real-world benchmark datasets show that for both methods, there is a statistically significant correlation between the feasibility of the measure, and the actual improvement in classification accuracy of the whole classifier system; therefore, both feasibility measures can be used in practical applications to choose an optimal classification confidence measure.

- [3]
**Prokop L., Mišák S., Snášel V., Platos J., Krömer P.:**

Supervised learning of photovoltaic power plant output prediction models, 321-338.

**First page Full text DOI: 10.14311/NNW.2013.23.020**

This article presents an application of evolutionary fuzzy rules to the modeling and prediction of power output of a real-world Photovoltaic Power Plant (PVPP). The method is compared to artificial neural networks and support vector regression that were also used to build predictors in order to analyse a time-series like data describing the production of the PVPP. The models of the PVPP are created using different supervised machine learning methods in order to forecast the short-term output of the power plant and compare the accuracy of the prediction.

- [4]
**Markechová D.:**

Entropy and mutual information of experiments in the fuzzy case, 339-350.

**First page Full text DOI: 10.14311/NNW.2013.23.021**

In my previous papers ([18], [19]) the entropy of fuzzy partitions had been defined. The concept of the entropy of a fuzzy partition was used to define the entropy of a fuzzy dynamical system and to propose an ergodic theory for fuzzy dynamical systems ([19], [20]). In this paper, using my previous results related to the entropy of fuzzy partitions, a measure of average mutual information of fuzzy partitions is defined. Some properties concerning this measure are proved. It is shown that the entropy of fuzzy partitions can be considered as a special case of their mutual information. We obtain that subadditivity and additivity of entropy of fuzzy partitions are simple consequences of these properties. The suggested measures can be applied whenever it is need to know the amount of information that we obtain by realization of experiments, the results of which are fuzzy events.

- [5]
**Khalaj G., Pouraliakbar H., Mamaghani K. R., Khalaj M.-J.:**

Modeling the correlation between heat treatment, chemical composition and bainite fraction of pipeline steels by means of artifcial neural networks, 351-368.

**First page Full text DOI: 10.14311/NNW.2013.23.022**

In the present study, bainite fraction results of continuous cooling of high strength low alloy steels have been modeled by artificial neural networks. The artificial neural network models were constructed by 16 input parameters including chemical compositions (C, Mn, Nb, Mo, Ti, N, Cu, P, S, Si, Al, V), Nb in solution, austenitizing temperature, initial austenite grain size and cooling rate over the temperature range of the occurrence of phase transformations. The value for the output layer was the bainite fraction. According to the input parameters in feed-forward back-propagation algorithm, the constructed networks were trained, validated and tested. To make a decision on the completion of the training processes, two termination states are declared: state 1 (ANN-I model) means that the training of neural network was ended when the maximum epoch of process reached (1000) while state 2 (ANN-II model) means the training ended when minimum error norm of network gained. The entire statistical evaluators of ANN-II model has higher performance than those of ANN-I. However, both of the models exhibit valuable results and the entire statistical values show that the proposed ANN-I and ANN-II models are suitably trained and can predict the bainite fraction values very close to the experimental ones.

- [6]
**Mu-Yen Chen, Min-Hsuan Fan, Young-Long Chen, Hui-Mei Wei :**

Design of experiments on neural network's parameters optimization for time series forecasting in stock markets, 369-393.

**First page Full text DOI: 10.14311/NNW.2013.23.023**

Artificial neural network (ANN) model has been used for years to conduct research in stock price prediction for three reasons. First, it has a higher prediction accuracy rate in empirical research. Second, it is not subject to the assumption of having samples from a normal distribution. Third, it can deal with non-linear problems. Nevertheless, the accuracy of prediction relies on the parameter settings of neural network as well as the complexities of problems and the neural network architecture; the results of the analysis could be even more significant with the selection of optimal parameters and network architecture. Currently, as a way of setting parameters, most researchers employed the trial and error method. However, this method is very time-consuming and labor-intensive and may not result in the optimal parameters. Therefore, this research took advantage of a back propagation neural network (BPNN) for the purpose of parameter optimization through constructing a model of stock price prediction, applying design of experiment (DOE) to systematize experiment scheduling, and methods of main effects analysis and interaction analysis. The research used two datasets of financial ratios from 50 blue chip companies in Taiwanese stock market and 40 listed American banks in New York stock exchange as experimental samples. Research results showed that the correlation forecasting, root mean squared error (RMSE), and computing time, which can effectively increase the accuracy of stock price prediction, are better than traditional statistical methods and conventional neural network model.

### 3/2013

- [1]
**Maheshkumar Y., Ravi V., Abraham A. (India, USA):**

A particle swarm optimization-threshold accepting hybrid algorithm for unconstrained optimization, 191-221.

**First page Full text DOI: 10.14311/NNW.2013.23.013**

In this paper, we propose a novel hybrid metaheuristic algorithm, which integrates a Threshold Accepting algorithm (TA) with a traditional Particle Swarm Optimization (PSO) algorithm. We used the TA as a catalyst in speeding up convergence of PSO towards the optimal solution. In this hybrid, at the end of every iteration of PSO, the TA is invoked probabilistically to refine the worst particle that lags in the race of finding the solution for that iteration. Consequently the worst particle will be refined in the next iteration. The robustness of the proposed approach has been tested on 34 unconstrained optimization problems taken from the literature. The proposed hybrid demonstrates superior preference in terms of functional evaluations and success rate for 30 simulations conducted.

- [2]
**Škrinárová J., Huraj L., Siládi V. (Slovak Republic):**

A neural tree model for classification of computing grid resources using PSO tasks scheduling, 223-241.

**First page Full text DOI: 10.14311/NNW.2013.23.014**

This paper proposes a model of neural tree architecture with probabilistic neurons. These trees are used for classification of a large amount of computer grid resources to classes. The first tree is used for classification of hardware part of dataset. The second tree classifies patterns of software identifiers. Trees are implemented to successfully separate inputs into nine classes of resources. We propose Particle Swarm Optimization model for tasks scheduling in computer grid. We compared time of creation of schedule and time of makespan in six series of experiments without and with using neural trees. In experiments with using neural tree we gained the subset of suitable computational resources. The aim is effective mapping of a large batch of tasks into particular resources. On the base of experiments we can say that improvements have been made even for middle and small batch of tasks.

- [3]
**Lei Wang, Shenquan Liu, Yanjun Zeng (China):**

Diversity of firing patterns in a two-compartment model neuron: using internal time delay as an independent variable, 243-254.

**First page Full text DOI: 10.14311/NNW.2013.23.015**

Firing properties of single neurons in the nervous system have been recognized to be determined by their intrinsic ion channel dynamics and extrinsic synaptic inputs. Previous studies have suggested that dendritic structures exhibit significant roles in the modulation of somatic firing behavior in neurons. Following these studies, we show that finite information transmission delay between dendrite and soma can also influence the somatic firings in neurons. Our investigation is based on a two-compartment model which can approximately reproduce the firing activity of cortical pyramidal neurons. The obtained simulation results indicate that under subthreshold stimulus, spontaneous fast spiking activity is induced by large values of time delay, while for suprathreshold stimulus, regular bursting, chaotic firing and fast spiking can be observed under different time delays. More importantly, the transition mode between these diverse firing patterns with the variation of delay shows a period-doubling phenomenon under certain stimulus intensity. Consequently, our model results can not only illustrate the influential roles of internal time delay in the generation of a diversity of neuronal firing patterns, but also provide us with frameworks for investigating the impacts of internal time delay on the firing properties of many other neurons in the nervous system.

- [4]
**Peyghami M. R., Khanduzi R. (Iran):**

Novel MLP neural network with hybrid tabu search algorithm, 255-270.

**First page Full text DOI: 10.14311/NNW.2013.23.016**

In this paper, we propose a new global and fast Multilayer Perceptron Neural Network (MLP-NN) which can be used to forecast the automotive price. Nowadays, the gradient-based techniques, such as back propagation, are widely used for training neural networks. These techniques have local convergence results and, therefore, can perform poorly even on simple problems when forecasting is out of sample. On the other hand, the global search algorithms, like Tabu Search (TS), suffer from low rate convergence. Motivated by these facts, a new global and fast hybrid algorithm for training MLP-NN is provided. In our new framework, a hybridization of an extended version of TS with some local techniques is constructed in order to train the connected weights of the network. The extended version of TS in the proposed scheme consists of a simple TS together with the intensification and diversification search methods, and the local search methods are based on a direct strategy of Nelder-Mead (NM) or Levenberg-Marquardt (LM) techniques. This hybridization leads us to have a global and fast trained network in order to use in some forecasting problems. To show the efficiency and effectiveness of our new proposed network, we apply our new scheme for forecasting the automotive price in Iran Khodro Company which is the biggest car manufacturer in Iran. The results are promising compared to the cases when we apply the TS and some other forecasting techniques individually. We also compare the results with the case when we employ the gradient-based optimization techniques such as LM, and global search methods such as Genetic Algorithm (GA) and hybrid of MLP-NN with GA.

- [5]
**Garlík B., Křivan M. (Czech Republic):**

Identification of type daily diagrams of electric consumption based on cluster analysis of multi-dimensional data by neural network, 271-283.

**First page Full text DOI: 10.14311/NNW.2013.23.017**

This article establishes a mathematical description of a self-organizing neural network used for cluster analysis with a subsequent sampling of its effectiveness as an example of identification of the type daily diagrams of electric energy-consumption of complex intelligent buildings within an electric micro grid, namely for a typical work day and a day off on the basis of its annual history. The mentioned type daily diagram can be used to predict power consumption. This method is given in the context of the commonly used procedure for cluster analysis. The experiment was processed in the computer program Artint © 2010.

### 2/2013

- [1]
**Editorial, 79.**

- [2]
**Langer M., Kelemenová A. (Czech Republic):**

On positioned eco-grammar systems and pure grammars of type 0, 81-91.

**First page Full text DOI: 10.14311/NNW.2013.23.006**

In this paper we extend our results given in [5] where we compared PEG systems with pure regulated context-free grammars (see [3]). We will show that the family of languages generated by the pure grammars of type 0 is a proper subclass of the family of languages generated by positioned eco-grammar systems. We present a way how to coordinate parallel behavior of agents with one-sided context in a PEG system in order to simulate the derivation step in a pure grammar of type 0 determined by a single rule which replaces an arbitrarily long string by another one. Related results concerning PEG systems and pure languages can be found in [6].

- [3]
**Tuček P., Janoška Z. (Czech Republic):**

Fractal dimension as a descriptor of urban growth dynamics, 93-102.

**First page Full text DOI: 10.14311/NNW.2013.23.007**

The objective of this paper is to examine the development of the urban form of the city of Olomouc since the 1920s in terms of fractal dimension, and to link the observation with two other descriptors of shape - area and perimeter. The fractal dimension of built-up areas and fractal dimension of the boundary of the city are calculated employing the box-counting method; the possibilities of their interpretation and usage in urban planning are discussed. The process of urban growth is observed with respect to its fractality and perspectives of this approach are discussed. An interesting dependence between area and its fractal dimension is derived.

- [4]
**Balara D., Timko J., Žilková J. (Slovak Republic):**

Application of neural network model for parameters identification of non-linear dynamic system, 103-116.

**First page Full text DOI: 10.14311/NNW.2013.23.008**

A method for identification of parameters of a non-linear dynamic system, such as an induction motor with saturation effect taken into account, is presented in this paper. Adaptive identifier with structure similar to model of the system performs identification. This identifier can be regarded as a special neural network, therefore its adaptation is based on the gradient descent method and Back-Propagation well known in the neural networks theory. Parameters of electromagnetic subsystems were derived from the values of synaptic weights of the estimator after its adaptation. Testing was performed with simulations taking into account noise in measured quantities. Deviations of identified parameters in case of electrical parameters of the system were up to 1% of real values. Parameters of non-linear magnetizing curve were identified with deviations up to 6% of real values. Identifier was able to follow sudden changes of rotor resistance, load torque and moment of inertia.

- [5]
**Khalaj G., Nazari A., Pouraliakbar H. (Iran):**

Prediction of martensite fraction of microalloyed steel by artificial neural networks, 117-130.

**First page Full text DOI: 10.14311/NNW.2013.23.009**

The final microstructure and resulting mechanical properties in the linepipe steels are predominantly determined by austenite decomposition during cooling after thermomechanical and welding processes. The paper presents some results of the research connected with the development of a new approach based on the artificial neural network to predicting the martensite fraction of the phase constituents occurring in five microalloyed steels after continuous cooling. The independent variables in the model are chemical compositions, niobium condition, austenitizing temperature, initial austenite grain size and cooling rate over the temperature range of the occurrence of phase transformations. For the purpose of constructing these models, 104 different experimental data were gathered from the literature. According to the input parameters in feedforward backpropagation algorithm, the constructed networks were trained, validated and tested. In this model, the training and testing results in the artificial neural network have shown a strong potential for prediction of effects of chemical compositions and heat treatments on phase transformation of microalloyed steels.

- [6]
**Martinovič J., Slaninová K., Vojáček L., Dráždilová P., Dvorský J., Vondrák I.(Czech Republic):**

Effective clustering algorithm for high-dimensional sparse data based on SOM, 131-147.

**First page Full text DOI: 10.14311/NNW.2013.23.010**With increasing opportunities for analyzing large data sources, we have noticed a lack of effective processing in datamining tasks working with large sparse datasets of high dimensions. This work focuses on this issue and on effective clustering using models of artificial intelligence.

The authors of this article propose an effective clustering algorithm to exploit the features of neural networks, and especially Self Organizing Maps (SOM), for the reduction of data dimensionality. The issue of computational complexity is resolved by using a parallelization of the standard SOM algorithm. The authors have focused on the acceleration of the presented algorithm using a version suitable for data collections with a certain level of sparsity. Effective acceleration is achieved by improving the winning neuron finding phase and the weight actualization phase. The output presented here demonstrates sufficient acceleration of the standard SOM algorithm while preserving the appropriate accuracy.

- [7]
**Anitha J., Jude Hemanth D. (India):**

An efficient Kohonen-fuzzy neural network based abnormal retinal image classification system, 149-167.

**First page Full text DOI: 10.14311/NNW.2013.23.011**

Artificial Neural Network (ANN) is the primary automated AI system preferred for medical applications. Even though ANN possesses multiple advantages, the convergence of the ANN is not always guaranteed for the practical applications. This often results in the local minima problem and ultimately yields inaccurate results. This convergence problem is common among ANNs and especially in Kohonen neural networks which employ unsupervised training methodology. In this work, an Efficient Kohonen Fuzzy Neural (EKFN) network is proposed to eliminate the iteration dependent nature of the conventional system. The suitability of this hybrid automated system is illustrated in the context of pathology identification in retinal images. This disease identification system includes anatomical structure segmentation from retinal images followed by image classification. The performance measures used are accuracy, sensitivity, specificity, positive predictive value and positive likelihood ratio. Experimental results show promising possibilities for the hybrid systems in terms of performance measures.

- [8]
**Lei Chen, Geng Yang, Yingzhou Zhang, Chuandong Wang, Zhen Yang (China):**

Asymptotically stable multi-valued many-to-many associative memory neural network and its application in image retrieval, 169-189.

**First page Full text DOI: 10.14311/NNW.2013.23.012**

As an important artificial neural network, associative memory model can be employed to mimic human thinking and machine intelligence. In this paper, first, a multi-valued many-to-many Gaussian associative memory model (M

^{3}GAM) is proposed by introducing the Gaussian unidirectional associative memory model (GUAM) and Gaussian bidirectional associative memory model (GBAM) into Hattori {et al}'s multi-module associative memory model ((MMA)^{2}). Second, the M^{3}GAM's asymptotical stability is proved theoretically in both synchronous and asynchronous update modes, which ensures that the stored patterns become the M^{3}GAM's stable points. Third, by substituting the general similarity metric for the negative squared Euclidean distance in M^{3}GAM, the generalized multi-valued many-to-many Gaussian associative memory model (GM^{3}GAM) is presented, which makes the M^{3}GAM become its special case. Finally, we investigate the M^{3}GAM's application in association-based image retrieval, and the computer simulation results verify the M^{3}GAM's robust performance.

### 1/2013

- [1]
**Editorial, 1-2.**

- [2]
**Poláková R., Tvrdík J. (Czech Republic):**

A combined approach to adaptive differential evolution, 3-15.

**First page Full text DOI: 10.14311/NNW.2013.23.001**

The paper deals with the adaptive mechanisms in differential evolution (DE) algorithm. DE is a simple and effective stochastic algorithm frequently used in solving the real-world global optimization problems. The efficiency of the algorithm is sensitive to setting its control parameters. Several adaptive approaches have appeared recently in order to avoid control-parameter tuning. A new adaptive variant of differential evolution is proposed in this study. It is based on a combination of two adaptive approaches published before. The new algorithm was tested on the well-known set of benchmark problems developed for the special session of CEC2005 at four levels of population size and its performance was compared with the adaptive variants that were applied in the design of the new algorithm. The new adaptive DE variant outperformed the others in several test problems but its efficiency on average was not better.

- [3]
**Bujok P. (Czech Republic):**

Synchronous and asynchronous migration in adaptive differential evolution algorithms, 17-30.

**First page Full text DOI: 10.14311/NNW.2013.23.002**

The influence of synchronous and asynchronous migration on the performance of adaptive differential evolution algorithms is investigated. Six adaptive differential evolution variants are employed by the parallel migration model with a~star topology. Synchronous and asynchronous migration models with various parameters settings were experimentally compared with non-parallel adaptive algorithms in six shifted benchmark problems of dimension

*D*= 30. Three different ways of exchanging individuals are applied in a~synchronous island model with a fixed number of islands. Three different numbers of sub-populations are set up in an asynchronous island model. The parallel synchronous and asynchronous migration models increase performance in most problems.

- [4]
**Sosík P., Patón A., Ciencialová L. (Czech Republic):**

Polynomial time-bounded computations in spiking neural P systems, 31-48.

**First page Full text DOI: 10.14311/NNW.2013.23.003**

The paper introduces a formal framework for the study of computational power of spiking neural (SN) P systems. We define complexity classes of uniform families of recognizer SN P systems with and without input, in a way which is standard in P systems theory. Then we study properties of the resulting complexity classes, extending previous results on SN P systems. We demonstrate that the computational power of several variants of confluent SN P systems, under polynomial time restriction, is characterized by classes ranging from

**P**to**PSPACE**.

- [5]
**Lenčuchová J. (Slovak Republic):**

Testing MSW type of nonlinearity using autocopulas, 49-60.

**First page Full text DOI: 10.14311/NNW.2013.23.004**

Inspired by Rakonczai et al. [8], we use autocopulas for the testing of linearity against Markov-switching type of nonlinearity and remaining nonlinearity. They applied this autocopula approach to testing heteroscedasticity in AR-ARCH model. Given a strictly stationary time series

*Y*, a_{t}*k*-lag autocopula is a bivariate joint distribution function of the random vector (*Y*). Our contribution is in extending the idea to test the linearity against Markov-switching type of nonlinearity and remaining nonlinearity [5,6] in order to avoid classical, time-consuming tests._{t}, Y_{t-k}

- [6]
**Stoklasa J., Jandová V., Talašová J. (Czech Republic):**

Weak consistency in Saaty's AHP - evaluating creative work outcomes of Czech Art Colleges, 61-77.

**First page Full text DOI: 10.14311/NNW.2013.23.005**

The full consistency of Saaty's matrix of preference intensities used in AHP is practically unachievable for a large number of objects being compared. There are many procedures and methods published in the literature that describe how to assess whether Saaty's matrix is "consistent enough". Consistency is in these cases measured for an already defined matrix (i.e. ex-post). In this paper we present a procedure that guarantees that an acceptable level of consistency of expert information concerning preferences will be achieved. The proposed method is based on dividing the process of inputting Saaty's matrix into two steps. First, the ordering of the compared objects with respect to their significance is determined using the pairwise comparison method. Second, the intensities of preferences are defined for the objects numbered in accordance with their ordering (resulting from the first step). In this paper the weak consistency of Saaty's matrix is defined, which is easy to check during the process of inputting the preference intensities. Several propositions concerning the properties of weakly consistent Saaty's matrices are proven in the paper. We show on an example that the weak consistency, which represents a very natural requirement on Saaty's matrix of preference intensities, is not achieved for some matrices, which are considered "consistent enough" according to the criteria published in the literature. The proposed method of setting Saaty's matrix of preference intensities was used in the model for determining scores for particular categories of artistic production, which is an integral part of the Registry of Artistic Results (RUV) currently being developed in the Czech Republic. The Registry contains data on works of art originating from creative activities of Czech art colleges and faculties. Based on the total scores achieved by these institutions, a part of the state budget subsidy is being allocated among them.