Contents of Volume 19 (2009)

1/2009 2/2009 3/2009 4/2009 5/2009


  • [1] Kala R., Shukla A., Tiwari R. (India): Self-adaptive parallel processing neural networks with irregular nodal processing powers using hierarchical partitioning, 657-680.

    The architecture and working of the Artificial Neural Networks are an inspiration from the human brain. The brain due to its highly parallel nature and immense computational powers still remains the motivation for researchers. A single system-single processor approach is a highly unlikely way to model a neural network for large computational needs. Many approaches have been proposed that adopt a parallel implementation of ANNs. These methods do not consider the difference in processing powers of the constituting units and hence workload distribution among the nodes is not optimal. Human brain not always has equal processing power among the neurons. A person having disability in some part of brain may be able to perform every task with reduced capabilities. Disabilities weaken the processing of some parts. This inspires us to make a self-adaptive system of ANN that would optimally distribute computation among the nodes. The self-adaptive nature of the algorithm makes it possible for the algorithm to taper dynamic changes in node performance. We used data, node and layer partitioning in a hierarchical manner in order to evolve the most optimal architecture comprising of the best features of these partitioning techniques. The adaptive hierarchical architecture enables performance optimisation in whatever condition and problem the algorithm is used. The system was implemented and tested on 20 systems working in parallel. Besides, the computational speed-up, the algorithm was able to monitor changes in performance and adapt accordingly.

  • [2] Zhiyi Yin, Fuxi Zhu, Jianming Fu, Debin Gao (China, Singapore): Finding the same source programs based on the structural fingerprint distance of call graph, 681-693.

    With the purpose of guaranteeing the copyright and security of software, we introduce the structural fingerprint and the distance of the fingerprint to find the same source programs from a great deal of programs in this paper. In order to gain the structural fingerprint, the in-degree, out-degree and adjacency relationship are exacted from call graph to construct a structural matrix. Then this matrix is mapped to RGB image and to compute the color moments of this image. Comparing with the traditional binary comparison way in which finding graph isomorphism is based on control flow graph or instruction similarity, this method offers many advantages in application. First of all, the image processing techniques are made full use of to gain the color moments that are considered as the structural fingerprint to identify different programs. And secondly, the distance of structural fingerprint can be used to find the same source programs from a large number of programs. Last but not least, the runtime of our method is significantly shorter than the traditional methods. It takes only seconds comparing to minutes and even hours taken by other methods.

  • [3] Izakian H., Abraham A., Snášel V. (Czech Republic): Performance comparison of six efficient pure heuristics for scheduling meta-tasks on heterogeneous distributed environments, 695-710.

    Scheduling is one of the core steps to efficiently exploit the capabilities of heterogeneous distributed computing systems and represents an NP-complete problem. Therefore, using meta-heuristic algorithms is a suitable approach in order to cope with its difficulty. In many meta-heuristic algorithms, generating individuals in the initial step has an important effect on the convergence behavior of the algorithm and final solutions. Using some pure heuristics for generating one or more near-optimal individuals in the initial step can improve the final solutions obtained by meta-heuristic algorithms. Pure heuristics may be used solitary for generating schedules in many real-world situations in which using the meta-heuristic methods are too difficult or inappropriate. Different criteria can be used for evaluating the efficiency of scheduling algorithms, the most important of which are makespan and flowtime. In this paper, we propose an efficient pure heuristic method and then we compare the performance with five popular heuristics for minimizing makespan and flowtime in heterogeneous distributed computing systems. We investigate the effect of these pure heuristics for initializing simulated annealing meta-heuristic approach for scheduling tasks on heterogeneous environments.

  • [4] Zhao Rongchang, Ma Yide, Zhan Kun (China): Tri-state cascading pulse coupled neural network and its application in finding shortest path, 711-723.

    To increase the computing speed of neural networks by means of parallel performance, a new mode of neural network, named Tri-state Cascading Pulse Coupled Neural Network (TCPCNN), is presented in this paper, which takes the ideas of three-state and pipelining used in circuit designing into neural network, and creates new neuron with three states: sub-firing, firing and inhibition. The proposed model can transmit signals in parallel way, as it is inspired not only in the direction of auto-wave propagation but also in its transverse direction in neural network. In this paper, TCPCNN is applied to find the shortest path, and the experimental results indicate that the algorithm has lower computational complexity, higher accuracy, and secured full-scale searching. Furthermore, it has little dependence on initial conditions and parameters. The algorithm is tested by some experiments, and its results are compared with some other classical algorithms --- Dijkstra algorithm, Bellman-Ford algorithm and a new algorithm using pulse coupled neural networks.

  • [5] Dűndar P., Kılıc E., Balcı M. A. (Turkey): Finding the path centre of a communications network, 725-733.

    The centre of a communications network is a vertex set. The distances between every vertex in the centre set and all other vertices of the network are minimal. In some cases, the centre of the network can be a path, which includes a desired number of vertices. This centre is called a path centre of the network. In this paper, we aim to find a path centre of a given network with the needed number of vertices. We give the distance measures of the network and represent an algorithm searching the path centre of the network.

  • [6] Walendziak A., Wojciechowska-Rysiawa M. (Poland): Another axiomatization of pseudo-BL algebras, 735-743.

    A system of axioms defining a pseudo-BL algebra is given.

  • [7] Radoň T. (Czech Republic): Model of MAC sub-layer of a CAN protocol by using hierarchical Coloured Petri Net, 745-761.

    This article presents a design of model of the medium access control (MAC) of a sub-layer of a Controller Area Network (CAN) protocol (CAN is the most widely used in-vehicle network). The model is created via hierarchical Coloured Petri Nets. For better clarity and comprehension, the wide created model is divided into submodules.

    An application CPN Tools, developed by the CPN group at the University of Aarhus (Denmark), is used as a modelling tool. This model expresses the whole CAN's fault confinement mechanisms and the other functions of MAC sub-layer such as data encapsulation, frame coding (stuffing/de-stuffing), medium access management and acknowledgement. Functionality of the originally created model was tested by a series of ad hoc simulations in the model environment. The assets of the model mentioned before are discussed at the end of the article.

  • [8] Contents volume 19 (2009), 763-765.
  • [9] Author's index volume 19 (2009), 767-770.


  • [1] Editorial, 399-400.
  • [2] Beliczynski B., Ribeiro B. (Poland, Portugal): Some enhancements to approximation of one-variable functions by orthonormal basis, 401-412.

    Some enhancements to the approximation of one-variable functions with respect to an orthogonal basis are considered. A two-step approximation scheme is presented here. In the first step, a constant bias is extracted from the approximated function, while in the second, the function with extracted bias is approximated in a usual way. Later, these two components are added together. First of all we prove that a constant bias extracted from the function decreases the error. We demonstrate how to calculate that bias. Secondly, in a minor contribution, we show how to choose basis from a selected set of orthonormal functions to achieve minimum error. Finally we prove that loss of orthonormality due to truncation of the argument range of the basis functions does not effect the overall error of approximation and the expansion coefficients' correctness. We show how this feature can be used. Our attention is focused on Hermite orthonormal functions. An application of the obtained results to ECG data compression is presented.

  • [3] Érdi P., Ujfalussy B., Diwadkar V. (Hungary, USA): The schizophrenic brain: A broken hermeneutic circle, 413-427.

    A unifying picture to the hermeneutical approach to schizophrenia is given by combining the philosophical and the experimental/computational approaches. Computational models of associative learning and recall in the cortico-hippocampal system helps to understand the circuits of normal and pathological behavior.

  • [4] Borisyuk R., Chik D., Kazanovich Y. (USA, Russia): Selective attention model of moving objects, 429-445.

    Tracking moving objects is a vital visual task for the survival of an animal. We describe oscillatory neural network models of visual attention with a central element that can track a moving target among a set of distracters on the screen. At the initial stage, the model forms the focus of attention on an arbitrary object that is considered as a target. Other objects are treated as distracters. We present here two models: 1) synchronisation based model designed as a network of phase oscillators and 2) spiking neural model which is based on the idea of resource-limited parallel visual pointers. Selective attention and the tracking process are represented by the partial synchronisation between the central unit and a subgroup of peripheral elements. Simulation results are in overall agreement with the findings from psychological experiments: overlapping between the target and distractors is the main source of errors.

  • [5] Hunter R., Cobb S., Graham B. P. (United Kingdom): Improving associative memory in a network of spiking neurons, 447-470.

    Associative neural network models are a commonly used methodology when investigating the theory of associative memory in the brain. Comparisons between the mammalian hippocampus and associative memory models of neural networks have been investigated [12]. Biologically based networks are systems built of complex biologically realistic cells with a variety of properties. Here we compare and contrast associative memory function in a network of biologically-based spiking neurons [22] with previously published results for a simple artificial neural network model [11]. We shall focus primarily on the recall process from a memory where patterns have previously been stored by Hebbian learning. We investigate biologically plausible implementations of methods for improving recall under biologically realistic conditions, such as a sparsely connected network. Network dynamics under recall conditions are further tested using network configurations including complex multi-compartment inhibitory interneurons, known as basket cells.

  • [6] Cutsuridis V., Cobb S., Graham B. P. (United Kingdom): Modeling the STDP symmetry-to-asymmetry transition in the presence of gabaergic inhibition, 471-481.

    Experimental studies have shown a symmetry-to-asymmetry transition of the spike-timing dependent plasticity (STDP) curve exists in the proximal stratum radiatum (SR) dendrite of the hippocampal CA1 pyramidal neuron, which is probably due to the presence of GABAergic inhibition [2, 3, 4]. A recent computational model predicted that symmetry-to-asymmetry transition is strongly dependent on the frequency and conductance value of GABA inhibition and that the largest long term potentiation (LTP) value and the two distinct long-term depression (LTD) tails of the symmetrical STDP curve are centred at +10 ms, +40 ms and -10 ms, respectively [8, 9]. In the present paper, we continue to investigate even further via computer simulations the effects of gamma frequency inhibition and its conductance value to the symmetry-to-asymmetry transition of the STDP profile in the SR dendrite and predict that the transition is even more robust when there is a temporal offset between the onsets of the pre-post excitatory stimulation and the GABAergic inhibition. The largest LTP value and the two distinct LTD tails are inversely proportional to the increase of GABA conductance.

  • [7] Totoki Y., Mitsunaga K., Suemitsu H., Matsuo T. (Japan): Firing pattern estimation and synchronization detection of synaptically coupled Hindmarsh-Rose neurons, 483-497.

    In this paper, we present adaptive observers for synaptically coupled Hindmarsh-Rose (HR) neurons with the membrane potential measurement under the assumption that some of parameters in an individual HR neuron are known. Using the adaptive observers for a single HR neuron, we propose a two-stage merging procedure to identify the firing pattern of a model of synaptically coupled HR neurons. The procedure allows us to recover the internal states and to distinguish the firing patterns of the synaptically coupled HR neurons, with early-time dynamic behaviors.

  • [8] Liu J., Erwin H., Wermter S. (United Kingdom): From the inferior colliculus to a computational sound localization model, 499-512.

    In this paper, we describe a spiking neural network for building an azimuthal sound localization system, which is inspired by the functional organization of the human auditory midbrain up to the inferior colliculus (IC). Our system models two ascending pathways from the cochlear nucleus to the IC: an ITD (Interaural Time Difference) pathway and an ILD (Interaural Level Difference) pathway. We take account of Yin's finding [1] that multiple delay lines only exist in the contralateral medial superior olive (MSO) in our modeling of the ITD pathway. A level-locking auditory neuron is introduced for the ILD pathway network to encode sound amplitude into spike sequences. At the IC level, we differentiate between a low frequency (below 1 kHz) and high frequency (above 1 kHz) sound when combining the ITD and ILD cues to compute the azimuth angle of a sound. This paper provides a detailed illustration of the biological evidence of our hybrid ITD and ILD model. Experimental results of several types of sound are presented to evaluate our system. (This paper is an extension to one [2] of our papers in 2008 International Conference on Artificial Neural Networks.)

  • [9] Szilágyi S., Szilágyi L., Iclănzan D., Dávid L., Frigy A., Benyó Z. (Romania, Hungary): Intensity inhomogeneity correction and segmentation of magnetic resonance images using a multi-stage fuzzy clustering approach, 513-528.

    Intensity inhomogeneity or intensity non-uniformity (INU) is an undesired phenomenon that represents the main obstacle for magnetic resonance (MR) image segmentation and registration methods. Various techniques have been proposed to eliminate or compensate the INU, most of which are embedded into clustering algorithms, and they generally have difficulties when INU reaches high amplitudes. This study proposes a multiple stage fuzzy c-means (FCM) clustering based algorithm for the estimation and compensation of INU, by modelling it as a slowly varying additive or multiplicative noise. The slowly varying behaviour of the bias or gain field is assured by a smoothing filter that performs a context dependent averaging, controlled by a morphological criterion. The segmentation is also supported by a prefiltering technique for Gaussian and impulse noise elimination. The experiments using 2-D synthetic phantoms and real MR images indicate that the proposed method provides accurate segmentation. The resulting segmentation and fuzzy membership values can serve as excellent support for 3-D registration and surface reconstruction techniques.

  • [10] Sato Y. D., Jitsev J., von der Malsburg C. (Germany): A visual object recognition system invariant to scale and rotation, 529-544.

    We address here the problem of scale and rotation invariant object recognition, making use of a correspondence-based mechanism, in which the identity of an object represented by sensory signals is determined by matching it to a representation stored in memory. The sensory representation is in general affected by various transformations, notably scale and rotation, thus giving rise to the fundamental problem of invariant object recognition. We focus here on a neurally plausible mechanism that deals simultaneously with identification of the object and detection of the transformation, both types of information being important for visual processing. Our mechanism is based on macrocolumnar units. These evaluate identity- and transformation-specific feature similarities, performing competitive selection of the alternatives of their own subtask, and cooperate to make a coherent global decision for the identity, scale and rotation of the object.

  • [11] Weiler D., Clemente I. A., Willert V., Eggert J. (Germany): A probabilistic prediction method for object contour tracking, 545-560.

    We present an approach for probabilistic contour prediction within the framework of an object tracking system. We combine level-set methods for image segmentation with optical flow estimations based on probability distribution functions (pdfs) calculated at each image position. Unlike most recent level-set methods that consider exclusively the sign of the level-set function to determine an object and its background, we introduce a novel interpretation of the value of the level-set function that reflects the confidence in the contour. To this end, in a sequence of consecutive images, the contour of an object is transformed according to the optical flow estimation and used as the initial object hypothesis in the following image. The values of the initial level-set function are set according to the optical flow pdfs and thus provide an opportunity to incorporate the uncertainties of the optical flow estimation in the object contour prediction.

  • [12] Simou N., Athanasiadis T., Kollias S., Stamou G. (Greece): Semantic adaptation of neural network classifiers in image segmentation, 561-579.

    Semantic analysis of multimedia content is an ongoing research area that has gained a lot of attention over the last few years. Additionally, machine learning techniques are widely used for multimedia analysis with great success. This work presents a combined approach aiming at the semantic adaptation of neural network classifiers in a multimedia framework. Our proposal is based on a fuzzy reasoning engine which is able to evaluate the outputs and the confidence levels of the neural network classifier, using a domain specific knowledge base. The results obtained by the fuzzy reasoning engine are used as input for the adaptation of the network classifier, further increasing its ability to provide accurate classification of the specific content. The improved performance of the adapted neural network is used by a semantic segmentation algorithm that merges neighbouring regions satisfying certain criteria. In that way, fine image segmentation and classification are established.

  • [13] Labusch K., Barth E., Martinetz T. (Germany): Demixing jazz-music: Sparse coding neural gas for the separation of noisy overcomplete sources, 581-596.

    We consider the problem of separating noisy overcomplete sources from linear mixtures, i.e., we observe N mixtures of M > N sparse sources. We show that the ``Sparse Coding Neural Gas'' (SCNG) algorithm [8,9] can be employed in order to estimate the mixing matrix. Based on the learned mixing matrix the sources are obtained by orthogonal matching pursuit. Using synthetically generated data, we evaluate the influence of (i) the coherence of the mixing matrix, (ii) the noise level, and (iii) the sparseness of the sources with respect to the performance that can be achieved on the representation level. Our results show that if the coherence of the mixing matrix and the noise level are sufficiently small and the underlying sources are sufficiently sparse, the sources can be estimated from the observed mixtures. In order to apply our method to real-world data, we try to reconstruct each single instrument of a jazz audio signal given only a two-channel recording. Furthermore, we compare our method to the well-known FastICA [4] algorithm and show that in case of sparse sources and presence of additive noise, our method provides a superior estimation of the mixing matrix.

  • [14] Otsuka M., Yoshimoto J., Doya K. (Japan): Reward-dependent sensory coding in free-energy-based reinforcement learning, 597-610.

    The free-energy-based reinforcement learning is a new approach to handling high-dimensional states and actions. We investigate its properties using a new experimental platform called the digit floor task. In this task, the high-dimensional pixel data of hand-written digits were directly used as sensory inputs to the reinforcement learning agent. The simulation results showed the robustness of the free-energy-based reinforcement learning method against noise applied in both the training and testing phases. In addition, reward-dependent sensory representations were found in the distributed activation patterns of hidden units. The representations coded in a distributed fashion persisted even when the number of hidden nodes were varied.

  • [15] Ke Sun, Lei Xu (Hong Kong): Bayesian Ying-Yang learning on orthogonal binary factor analysis, 611-624.

    Binary Factor Analysis (BFA) aims to discover latent binary structures in high dimensional data. Parameter learning in BFA faces an exponential computational complexity and a large number of local optima. The model selection to determine the latent binary dimension is therefore difficult. Traditionally, it is implemented in two separate stages with two different objectives. First, parameter learning is performed for each candidate model scale to maximise the likelihood; then the optimal scale is selected to minimise a model selection criterion. Such a two-phase implementation suffers from huge computational cost and deteriorated learning performance on large scale structures. In contrast, the Bayesian Ying-Yang (BYY) harmony learning starts from a high dimensional model and automatically deducts the dimension during learning. This paper investigates model selection on a subclass of BFA called Orthogonal Binary Factor Analysis (OBFA). The Bayesian inference of the latent binary code is analytically solved, based on which a BYY machine is constructed. The harmony measure that serves as the objective function in BYY learning is more accurately estimated by recovering a regularisation term. Experimental comparison with the two-phase implementations shows superior performance of the proposed approach.

  • [16] Fiannaca A., Di Fatta G., Rizzo R., Urso A., Gaglio S. (Italy, United Kingdom): Clustering quality and topology preservation in fast learning SOMs, 625-639.

    The Self-Organizing Map (SOM) is a popular unsupervised neural network able to provide effective clustering and data visualization for data represented in multidimensional input spaces. In this paper, we describe Fast Learning SOM (FLSOM) which adopts a learning algorithm that improves the performance of the standard SOM with respect to the convergence time in the training phase. We show that FLSOM also improves the quality of the map by providing better clustering quality and topology preservation of multidimensional input data. Several tests have been carried out on different multidimensional datasets, which demonstrate better performances of the algorithm in comparison with the original SOM.

  • [17] Valenzuela W., Carvajal G., Figueroa M. (Chile): Blind source-separation in mixed-signal VLSI, 641-656.

    This paper describes an implementation of the Kurtosis and InfoMax algorithms for an independent components analysis in mixed-signal CMOS. Our design uses on-chip calibration techniques and local adaptation to compensate for the effect of device mismatch in arithmetic blocks and analog memory cells. We use our design to perform two-input blind source-separation on mixtures of audio signals and mixtures of EEG signals. Our experiments show that the hardware implementation of InfoMax consistently separates the signals within a normalized reconstruction error of less than 10%, while the reconstruction error of Kurtosis varies between 25% and 60%, due to its higher sensitivity to device mismatch and input statistics. Each circuit has a settling time of 8 ms, occupies a die area of 0.016-0.022 mm² and dissipates 15-20 mW of power.


  • [1] Editorial, 333-335.
  • [2] Hyland G. J. (Germany): Frőhlich's physical theory of cancer -- Frőhlich's path from theoretical physics to biology, and the cancer problem, 337-354.

    A review is given of Frőhlich's approach to biology from the side of theoretical physics, and illustrated in the context of his prediction of three types of coherent excitations in living systems based on their dielectric and elastic properties and far-from-equilibrium (non-linear) character. Supporting experimental evidence is presented, and the difficulty in achieving reproducibility addressed. His envisaged role of coherent excitations in cell division and its control is outlined, together with the implications for cancer - as understood at the time of his work.

  • [3] Del Giudice E., Elia V., Tedeschi A. (Italy): Role of water in the living organisms, 355-360.

    It is shown that coherent electrodynamics of water molecules produces extended regions where the chemical activity of bio-molecules is governed in a selective way by a code based on frequency resonance. Coherence Domains of water act as devices able to collect low-grade energy in the environment and to transform it into high-grade energy able to produce electronic excitations.

    Water is the most important constituent of all living organisms (70% of the total mass and 99% of all molecules). Other biomolecules, proteins, fats, sugars, vitamins, salts, which are usually considered the only molecules playing a remarkable role in molecular biology, make up only 1% of the total. So, biological activity is assumed to involve 1% of all molecules only.

    What is the role of water then? Is it possible that 99% of all biomolecules are necessary only as a solvent whereas the ``really essential'' biomolecules enact all productive activity?

    The driving and regulatory role of water in governing the biochemical activity has begun to be recognized in recent times (Voeikov, 2007).

    In order to unravel this puzzle, we should take another enigma, which is the existence of biochemical codes (Barbieri, 2004), into account. Apart from the living matter or more generally far from catalysts, molecules are usually subjected to a polygamous regime; each biomolecule can interact with many others, thus producing a great number of reactions. In living matter, instead, biomolecules live inside each particular biochemical cycle in a monogamous condition (at least within definite time intervals), i.e. a biomolecule interacts only with well-defined partners and ignores the other biomolecules, with which interaction would be possible in empty space. Living matter therefore produces a ``context'' capable of preventing a great number of chemical interactions, which would theoretically be possible. The possibility of molecular interactions is governed by biochemical codes (the genetic code is the most widely known among them), to which particular biological processes correspond. Within the world of biomolecules, there are thus the prerequisites for communication. Indeed, biochemical cycles are open and capable of reacting against new influences. In this way all the codes build up and adopt flexible features, which are typical of a language.

    The emergence of these biochemical codes from the dynamics of matter is undoubtedly the main problem of biology.

  • [4] Šrobár F. (Czech Republic): Role of non-linear interactions by the energy condensation in Frőhlich systems, 361-368.

    Properties of the Frőhlich model of collective oscillations of polar molecules (such as proteins) residing inside, or in the plasma membrane, of bio\-logical cells, receiving energy from intracellular metabolic sources and giving it off to surroundings, are studied. The exchange mechanisms with the heat bath can be linear and non-linear. Particular attention is given to condensation of energy in the lowest-frequency (fundamental) mode. Occurrence of this phenomenon, which is of importance for generation of electromagnetic fields by the cells as well as for other functions, presupposes that the non-linear interaction coefficient exceeds certain limit; if it is too small, the incoming energy is distributed to all modes. This can happen e.g. in the context of cancerous conditions due to perturbation of high static electric fields around mitochondria which are necessary support for non-linear behaviour. On the contrary, if interaction with the heat bath is too high, the oscillator system leaks energy to surroundings and must be pumped heavily to attain condensation.

  • [5] Pokorný J. (Czech Republic): Frőhlich's coherent vibrations in healthy and cancer cells, 369-378.

    Frőhlich formulated the hypothesis of coherent electrical polar oscillations in biological systems. The hypothesis predicts that generated electromagnetic field with a dominant electric component has a basic role in organization, transport, and interactions inside a cell and among cells. If mitochondria are entirely functional, the cellular cytoskeleton satisfies conditions for excitation of coherent states, which are assumed to be essential for normal biological activity. Malfunctioning mitochondria and disintegrated cytoskeleton result in disturbances of the Frőhlich's mechanism and consequently -- together with biochemical disturbances -- can lead to malignant properties of cancer cells.

  • [6] Jandová A., Pokorný J., Kobilková J., Trojan S., Nedbalová M., Dohnalová A., Čoček A., Mašata J., Holaj R., Tvrzická E., Zvolský P., Dvořáková M., Cifra M. (Czech Republic): Mitochondrial dysfunction, 379-391.

    Cell-mediated immunity (CMI) response of healthy humans and cancer (Ca) patients to specific tumor antigen and nonspecific (LDV -- lactate dehydrogenase virus) antigen, and of acute myocardial infarction (AMI) and schizophrenia (Sch) patients to nonspecific antigen was investigated. Large differences of CMI response of healthy humans in comparison with Ca, AMI, Sch patients were found. CMI response to antigens displays transferred information about cells under immune surveillance. LDV disturbs the oxidative energy production system. We assume that CMI response to LDV antigen monitors pathological states of mitochondrial energy production which results in disturbances of electromagnetic activity of living cells.

  • [7] Giuliani L., D'Emilia E., Lisi A., Grimaldi S., Foletti A., Del Giudice E. (Italy): The floating water bridge under strong electric potential, 393-398.

    Fuchs and collaborators [1, 2] showed that when a high voltage is applied between two electrodes, immersed in two beakers containing twice distilled water, a water bridge between the two containers is formed. We observed that a copper ions flow can pass through the bridge if the negative electrode is a copper electrode. The direction of the flux is not only depending on the direction of the applied electrostatic field but on the relative electronegativity of the electrodes too. The fact seems to suggest new perspectives in understanding the structure of water and the mechanisms concerning the arising of ions fluxes in living matter.


  • [1] Kahraman N., Erkmen B., Yıldırım T. (Turkey): Threshold voltage modeling using neural networks, 255-262.

    In this paper, threshold voltage modeling based on neural networks is presented. The database was obtained by performing DC analysis with possible combinations of MOSFETs terminal voltages and channel widths which directly effect threshold voltage values in submicron technology. The neural network was trained with the database including 0.25 ɲm and 0.40 ɲm TSMC process parameters. In order to prove the extrapolation ability, the test dataset is constituted with 0.18 ɲm TSMC process parameters, which were not applied to the neural network for training. The test results of neural network tool are compared with the data obtained by using the Cadence simulation tool. The excellent agreement between the experimental and the model results makes neural networks a powerful tool for estimation of the threshold voltage values.

  • [2] Popescu T. D. (Romania): Machine vibration monitoring by blind source separation and change detection, 263-277.

    The paper presents a new approach for a machine vibration analysis and health monitoring combining blind source separation (BSS) and change detection in source signals. So, the problem is translated from the space of the measurements to the space of independent sources, where the reduced number of components simplifies the monitoring problem and where the change detection methods are applied for scalar signals. The approach has been tested in simulation and the assessment on a real machine is presented in the last part of the paper.

  • [3] Karaboga D., Ozturk C. (Turkiye): Neural networks training by artificial bee colony algorithm on pattern classification, 279-292.

    Artificial Neural Networks are commonly used in pattern classification, function approximation, optimization, pattern matching, machine learning and associative memories. They are currently being an alternative to traditional statistical methods for mining data sets in order to classify data. Artificial Neural Networks are well-established technology for solving prediction and classification problems, using training and testing data to build a model. However, the success of the networks is highly dependent on the performance of the training process and hence the training algorithm. In this paper, we applied the Artificial Bee Colony (ABC) Optimization Algorithm on training feed-forward neural networks to classify different data sets which are widely used in the machine learning community. The performance of the ABC algorithm is investigated on benchmark classification problems from classification area and the results are compared with the other well-known conventional and evolutionary algorithms. The results indicate that ABC algorithm can efficiently be used on training feed-forward neural networks for the purpose of pattern classification.

  • [4] Rothkrantz L. J. M., Datcu D., Absil N. (Netherlands): Multimodal affect detection of car drivers, 293-305.

    The affective state of car drivers has a great impact on their driving behavior. In case of a high stress level, car drivers take more risk in turn over, car following and in neglecting traffic warning signs. In the paper we present first a list of factors which have an impact on the psychological state of the car driver with a focus on the affective state of car drivers. Next we give an outline of our multimodal system to assess the affective state of car drivers by analysis of facial expressions and speech signals. The output of our system is the basic input of a stress-o-meter providing the amount of the stress in a graphical way. We assume that this leads to a reduction of physical arousal and mental reappraisal of the situation. We present some experimental results of our assessment modules. So far we performed only lab experiments. At this moment we research noise canceling techniques in the car environment, which enable video and speech analysis in the car in real traffic environments.

  • [5] Svítek M. (Czech Republic): Quasi-non-ergodic probabilistic systems and wave probabilistic functions, 307-320.

    This paper presents models of quasi-non-ergodic probabilistic systems that are defined through the theory of wave probabilistic functions presented in [10-16]. First of all we show the new methodology on a binary non-ergodic time series. The theory is extended into $M$-dimensional non-ergodic $n$-valued systems with linear ergodicity evolution that are called quasi-non-ergodic probabilistic systems. We present two illustrative examples of applications of introduced theories and models.

  • [6] Yu B., Yang Z. Z., Yu B. (China): Hybrid model for multi-stop arrival time prediction, 321-332.

    Forecasting arrival times of a vehicle at many downstream stops is very important in many cases. For multi-stop arrival time prediction, direct approaches and iterative approaches possess respective merits. Therefore, a hybrid method that has both direct and iterative modeling abilities is presented to forecast arrival times at multiple stops. The hybrid method consists of an iterative support vector machine (SVM)-based prediction model and a direct SVM-based prediction model. In hybrid model, output from the iterative model is a rough prediction and it also needs to be adjusted, based on output from the direct model. The proposed model is assessed with the data of transit route number 3 in Guiyang city, China. Results show that the hybrid model seems to be a powerful tool for multi-stop arrival time prediction.


  • [1] El-Bakry H. M. (Egypt): A new neural design for faster pattern detection using cross correlation and matrix decomposition, 131-164.

    Neural networks have shown good results for detecting a certain pattern in a given image. In this paper, faster neural networks for pattern detection are presented. Such processors are designed based on cross correlation in the frequency domain between the input matrix and the input weights of neural networks. This approach is developed to reduce the computation steps required by these faster neural networks for the searching process. The principle of divide and conquer strategy is applied through matrix decomposition. Each matrix is divided into submatrices small in size, and then each one is tested separately by using a single faster neural processor. Furthermore, faster pattern detection is obtained by using parallel processing techniques to test the resulting submatrices at the same time, employing the same number of faster neural networks. In contrast to faster neural networks, the speed-up ratio is increased with the size of the input matrix when using faster neural networks and matrix decomposition. Moreover, the problem of local submatrix normalization in the frequency domain is solved. The effect of matrix normalization on the speed-up ratio of pattern detection is discussed. Simulation results show that local submatrix normalization through weight normalization is faster than submatrix normalization in the spatial domain. The overall speed-up ratio of the detection process is increased as the normalization of weights is done offline.

  • [2] Holeňa M., Steinfeldt N. (Czech Republic, Germany): Improving neural network approximations in applications: case study in materials science, 165-190.

    The popularity of feed-forward neural networks such as multilayer perceptrons and radial basis function networks is to a large extent due to their universal approximation capability. This paper concerns its theoretical principles, together with the influence of network architecture and of the distribution of training data on this capability. Then, the possibility to exploit this influence in order to improve the approximation capability of multilayer perceptrons by means of cross-validation and boosting is explained. Although in theory, the impact of both methods on the approximation capability of feed-forward networks is known, they are still not common in real-world applications. Therefore, the paper documents usefulness of both methods on a detailed case study in materials science.

  • [3] Kızılaslan R., Karlık B. (Turkey): Combination of neural networks forecasters for monthly natural gas consumption prediction, 191-199.

    This study presents different types of neural network algorithm based model forecasting gas consumption for residential and commercial consumers in Istanbul in Turkey. Using seven neural networks algorithms as forecasting models, we tried to find the best solution on forecasting of monthly natural gas consumption. These models were validated and tested on real monthly data from a distribution area covering two different regions of Anatolian and European sides in Istanbul. The analysis of results obtained for training and test sets show that the seven proposed artificial neural network models could be useful for the natural gas consumption forecast problem. It was shown that a conjugate gradient descent neural network model presented a more efficient solution than the other models.

  • [4] Vescan A. (Romania): Optimal component selection using a multiobjective evolutionary algorithm, 201-213.

    A component selection is a crucial problem in Component-Based Software Engineering (CBSE), which is concerned with the assembly of pre-existing software components.

    We are approaching the component selection involving dependencies between components. We formulate the problem as multiobjective, involving two objectives and one constraint. The approach used is an evolutionary computation technique. The experiments and comparisons with the greedy approach show the effectiveness of the proposed approach.

  • [5] Okatan A., Karlık B., Yağmur F. D. (Turkey): Detection of retinopathy diseases using artificial neural network based on discrete cosine transform, 215-221.

    The retinopathy diseases occur when the neurons do not transmit signals from retina to the brain. These disorders are: Diabetic retinopathy, hypertensive retinopathy, macular degeneration, vein branch occlusion, vitreous hemorrhage, and normal retina. This work presents a novel detection algorithm about retinopathy disorders from retina images. For this purpose, the retina images were pre-processed and resized at first. Then the discrete cosine transform was used as feature extraction before applying a neural network classifier. The performance of recognition rates of the novel detection algorithm were found as 50%, 70%, 85%, 90%, and 95% for testing five retinopathy cases respectively.

  • [6] Al-Shalfan K. A. (Saudi Arabia): Development of autonomous navigation wheelchairs based on fuzzy control, 223-233.

    Autonomous robotic wheelchairs based on visual guidance have been devoted to road edge detection. However, the after-detection process, especially the physical interpretation of what had been detected, needs more investigation. There is a wide gap between the scene model based on image processing algorithms and the physical model of the environment where the robotic wheelchair progresses. The aim of this paper is to investigate the interaction between the scene model and the world model; and also a visual control scheme for robot guidance that minimizes the model error induced by processing raw image data is proposed. This solution is developed based on a fuzzy control system, which uses the knowledge base information and the scene model to control the robot motion. On the other hand, the fuzzy control system is finely tuned by feed-backing the mean square errors between the scene model parameters and the knowledge-base data. Finally, the fuzzy controller uses results of these calculations to home the robot on the planned path. This paper also shows the principle of this system and the simulation results confirming the feasibility of the approach.

  • [7] Wiklendt L., Chalup S. T., Seron M. M. (Australia): Simulated 3d biped walking with an evolution-strategy tuned spiking neural network, 235-246.

    This paper presents the results of experiments in applying a spiking neural network to control the locomotion of a simulated biped robot. The neural model used in simulations was developed to allow for an analytic solution to a neuron fire time, while maintaining a non-instant post-synaptic potential rise time. The synaptic weights and delays were tuned using an evolution strategy. Simulation experiments demonstrate that within about seven thousand generations the biped is able to acquire a dynamic walk which allows it to walk upright for several metres.

  • [8] Grosan C., Abraham A. (Romania, USA): On a class of global optimization test functions, 247-252.

    This paper provides a theoretical proof illustrating that for a certain class of functions having the property that the partial derivatives have the same equation with respect to all variables, the optimum value (minimum or maximum) takes place at a point where all the variables have the same value. This information will help the researchers working with high dimensional functions to minimize the computational burden due to the fact that the search has to be performed only with respect to one variable.

  • [9] Notice, 253.


  • [1] Editorial, 1-2.
  • [2] Gutiérrez P. A., Hervás C., Fernández J. C., Jurado-Expósito M., Peña-Barragán J. M., López-Granados F. (Spain): Structural simplification of hybrid neuro-logistic regression models in multispectral analysis of remote sensed data, 3-20.

    Logistic Regression (LR) has become a widely used and accepted method to analyze binary or multiclass outcome variables, since it is a flexible tool that can predict probability for the state of a dichotomous variable. A recently proposed LR method is based on the hybridization of a linear model and Evolutionary Product Unit Neural Network (EPUNN) models for binary classification. This produces a high number of coefficients, so two different methods for simplifying the structure of the final model by reducing the number of initial or PU covariates are presented in this paper, both being based on the Wald test. The first method is a Backtracking Backward Search (BBS) method, and the other is similar, but it is based on the standard Simulated Annealing process for the decision steps (SABBS). In this study, we used aerial imagery taken in mid-May to evaluate the potential of two different combinations of LR and EPUNN (LR using PUs (LRPU), as well as LR using Initial covariates and PUs (LRIPU)) and the two presented methods of structural simplification of the final models (BBS and SABBS) used for discriminating Ridolfia segetum patches (one of the most dominant, competitive and persistent weed in sunflower crops) in a naturally infested field of southern Spain. Then, we compared the performance of these methods to six commonly used classification algorithms; our proposals obtaining a competitive performance and a lower number of coefficients.

  • [3] Fernandes B. J. T., Cavalcanti G. D. C., Ren T. I. (Brazil): Nonclassical receptive field inhibition applied to image segmentation, 21-36.

    This paper presents a new model to perform a supervised image segmentation task. The proposed model is called segmentation and classification with receptive fields (SCRF) which is based on the concept of receptive fields that analyzes pieces of an image considering not only a pixel or a group of pixels, but also the relationship between them and their neighbors. In order to work with the SCRF model, we propose a new artificial neural network, called I-PyraNet, which is a hybrid implementation of the recently described PyraNet and the nonclassical receptive fields inhibition. Furthermore, the model and the neural network are combined to accomplish a satellite image segmentation task.

  • [4] Salcedo Sanz S., Ortiz-García E. G., Pérez-Bellido Á. M., Portilla-Figueras A., Prieto L., Paredes D., Correoso F. (Spain): Performance comparison of multilayer perceptrons and support vector machines in a short-term wind speed prediction problem, 37-51.

    In this paper we present a comparison between the performance of Multilayer Perceptrons (MLPs) and Support Vector Machines (SVMs) in a problem of wind speed prediction. Specifically, we analyze the behavior of both algorithms within a larger system of wind speed prediction, formed by global and mesoscale weather forecasting models, and with a final statistical down-scaling process where the MLPs and the SVM are used. The final objective is to forecast the mean hourly wind speed prediction at wind turbines in a wind farm. This is an important parameter used to predict the total power production of the wind farm. The specific model for the short-term wind speed forecast we use integrates two different meteorological prediction global models, observations at the surface level and in different heights using atmospheric soundings. Also, it includes a mesoscale prediction model producing the inputs used in the MLP or the SVM, which will forecast the final wind speed at each turbine of the wind farm. In the experiments carried out we compare the results obtained using the MLP or SVM as final steps of the prediction system. Interesting differences of performance can be found when using MLPs or SVMs, which we analyze in this paper. The results obtained are encouraging anyway, and good short-term predictions of wind speed at specific points are obtained with both techniques.

  • [5] Gutiérrez J. M., Moreno-Barón L., del Valle M., Leija L., Muñoz R. (Spain, México): Wavelet neural network as a multivariate calibration method in voltammetric electronic tongues, 53-64.

    The paper presents a multi-output wavelet neural network (WNN) which, taking benefit of wavelets and neural networks, is able to accomplish data feature extraction and modeling. In this work, WNN is implemented with a feedforward one-hidden layer architecture, whose activation functions in its hidden layer neurons are wavelet functions, in our case, the first derivative of a Gaussian function. The network training is performed using a backpropagation algorithm, adjusting the connection weights along with the network parameters. This principle is applied to the simultaneous quantification of heavy metals present in liquid media, taking the cyclic voltammogram obtained with a modified epoxy-graphite composite sensor as departure information. The combination between processing tools and electrochemical sensors is already known as an electronic tongue.

  • [6] Krőmer P, Snášel V., Platoš J., Husek D. (Czech Republic): Genetic algorithms for the linear ordering problem, 65-80.

    Linear ordering problem is a well-known optimization problem attractive for its complexity (it is an NP-hard problem), rich library of test data and variety of real world applications. In this paper, we investigate the use and performance of two variants of genetic algorithms, mutation only genetic algorithms and higher level chromosome genetic algorithm, on the linear ordering problem. Both methods are tested and evaluated on a library of real world and artificial linear ordering problem instances.

  • [7] Parras-Gutierrez E., del Jesus M. J., Rivas V. M., Merelo J. J. (Spain): Study of the robustness of a meta-algorithm for the estimation of parameters in Radial Basis Function Neural Networks design, 81-84.

    Radial Basis Function Networks (RBFNs) have shown their capability to be used in classification problems, and therefore many data mining algorithms have been developed to configure RBFNs. These algorithms need to be given a suitable set of parameters for every problem they face, thus methods to automatically search the values of these parameters are required. This paper shows the robustness of a meta-algorithm developed to automatically establish the parameters needed to design RBFNs. Results show that this new method can be effectively used, not only to obtain good models, but also to find a stable set of parameters, available to be used on many different problems.

  • [8] Dostálová S., Šonka K. (Czech Republic): Untimely sleepiness: The problem and the current state of its management, 95-107.

    In modern society, many people keep working hours different from the standard (9 hours, 5 days a week during conventional time of the day), a regimen leading to sleep deprivation and/or circadian desynchronization, and, consequently, to sleepiness, fatigue, impaired efficiency and ultimately to psychic and somatic complaints. Fortunately, sleepiness-related health and occupational hazards can be kept under control using scientifically tested precautions designed to help maintain wakefulness. Suitable work and rest scheduling and observance of the principles of sleep hygiene are of major importance there. In situations which interfere with sleep, it is possible to use hypnotics or behavioral techniques and melatonin for circadian regimen optimization. In situations when sleep loss is temporarily inevitable, options to be taken into account include work shift shortening, breaks for rest, naps of short duration or administration of vigilance enhancing drugs. In the future, risks associated with sleepiness could be mitigated by means of currently developed technologies for real-time detection of sleepiness.

  • [9] Bouchner P., Faber J., Novotný S., Tichý T. (Czech Republic): Driver's attention level improvement with use of biofeedback stimulation incorporated into driving simulator, 109-118.

    The paper technically describes the principles of incorporation of the biofeedback system into the system of a driving simulator. After a brief introduction of the basic features of EEG biofeedback, the most important scenarios where such simulator enhancement can be successfully used are described. The system is introduced with the use of an analysis of the major technical and construction aspects, such as the software design, hardware realization and its incorporation into the driving simulator system. Finally, the paper sketches pilot experiments which were performed using EEG biofeedback incorporated into the driving simulator.

  • [10] Rothkrantz L. J. M., Horlings R., Dharmawan Z. (Netherlands): Recognition of emotional states of car drivers by EEG analysis, 119-128.

    It has been proved that the emotional state of a car driver has a great impact on his driving performance. Emotions can be triggered by internal stimuli (stress) or external stimuli (aggressive scenes). In this paper we propose a system of classification of certain emotional states by analysis of EEG recordings. We present the results of two experiments. In one experiment we recorded EEG data of car drivers in a simulated environment under conditions with a varying stress level. In the other experiment we presented pictures of emotional situations to car drivers. It proves that we were able to assess the state of respondents under extreme emotions.