Sections
Neuroinformatics - 2016
SESSION 1
Monday, April 25 17.15 – 18.30
Lecture-hall Алексеевский зал
Chair: Prof. DUNIN-BARKOWSKI WITALI
Neural networks and cognitive sciences
1. DUNIN-BARKOWSKI W.L., SOLOVYEVA K.P.1The Moscow Institute of Physics and Technology (State University)
2Scientific Research Institute for System Analysis, Moscow
Pavlov Principle in Brain Reverse Engineering
It is proposed a general principle of neural systems operating, which makes clear the cause of high efficiency of systems of neural-like elements - the Pavlov Principle (PP). The line of evolution of the ideas about neural mechanisms is followed beginning with Pavlov's discovery of systemic conditioned reflexes, through neuronal and synaptic schemes, early perceptrons, mid-time connectionism to modern deep learning constructions. A computational example of PP based model is given.
2. * SKITEVA LYUDMILA IGOREVNA, TROFIMOV ALEKSANDR GENNADIEVICH , USHAKOV VADIM LEONIDOVICH, VELICHKOVSKY BORIS MITROFANOVICH
1National Research Centre "Kurchatov Institute", Moscow
2National Research Nuclear University (MEPhI), Moscow
3
MEG data analysis using empirical mode decomposition
In the present paper it is proposed to use the empirical mode decomposition method for frequency bands analysis of MEG data, and the method is compared with the classical method of narrow band filtration and Hilbert transform. By the example of MEG data recorded during performing volitional sensorimotor activity by the subjects, it is shown that the procedure of extraction of empirical modes can detect useful information which is inaccessible to classical methods of frequency bands analysis.
3. POLEVAIA S.A., PARIN S.B., EREMIN E.V., BULANOV N.A., CHERNOVA M.A.
1N.I. Lobachevsky State University of Nizhni Novgorod
2Nizhny Novgorod State Medical Academy
3Higher School of Economics, Nizhny Novgorod Branch
Technology event-related telemetry for cognitive science
For the study of cognitive functions Introduced new technology: event-related telemetry and cognitive WEB-platform. The optimum architecture of information and telecommunication systems was choosed, to continuous monitoring of the functional state in the context of the sensorimotor activity in the management of feature of images in a virtual computer environment. Management errors considered as an objective display per-mary cognitive and sensorimotor coordination. It may be use for continious psychophysiological monitoring during daily life activity
4. VVEDENSKY V.L.
National Research Centre "Kurchatov Institute", Moscow
Self-organization of neuronal discharges in the cortex initiating action
With MEG we observe synchronous trains of beta-oscillations in different cortical areas before the start of self-paced finger movement. We conjecture, that these oscillations align neural trajectories in these cortical sites. This is the ordering process similar to formation of bird flocks, which brings the system of neurons into the state of self-organized criticality. System in this state can quickly react on relevant stimuli and ignore irrelevant ones. We discuss the role of alpha, beta and gamma rhythms in attention
5. EVGENII VITYAEV
Sobolev Institute of Mathematics, Siberian Branch of the Russian Academy of Sciences
COGNITOM FORMALIZATION
Cognitom formalization base on the formalization of functional systems cogs and phenomenological experience cogs by unified formalization of "natural" clasterization, "natural" concepts and integrated information by G.Tononi. This formalization base on the principal: the brain during the evolution process tuned on the extraction of the high correlational structure of the causal relations of the "natural" objects by discovery of "natural" clasterization and "natural" concepts.
POSTER SESSION 1
Tuesday, April 26 14.00 – 15.00
Lecture-hall Петровский зал
Chair: Prof. DOLENKO SERGEY
Applications of neural networks
6. IZOTOVA E.D., TARASOV D.S.1Kazan (Volga region) Federal University
2Meanotek, Kazan
Deep recurrent neural networks for extracting pharmacological terms from Russian texts
Recent studies have highlighted importance end-user reviews analysis for monitoring potential side effects of pharmaceuticals. In this work we apply deep recurrent neural networks, that recently have shown superior capabilities in term extraction tasks, for automated detection of important pharmacological terms in online reviews, posted by users of phramacological substances. We demonstrate that a many terms, can be reliably extracted this way using only a small (50k words) training dataset, and used for computing important statistics on drug usage. While accuracy of extraction of some other terms, such as perceived efficiency of the drugs was poor. We discuss possible strategies for improving our results and potential applications in pharmacovigilance.
7. BAKHSHIEV A.V., GUNDELAKH F.V.
1Peter the Great St. Petersburg Polytechnic University
2Russian State Scientific Center for Robotics and Technical Cybernetics, Saint-Petersburg
Modeling the motion memory mechanisms as element of robot's behavior control system
The article discusses the approach to controlling the behavior of robotic systems based on modeling the mechanisms of remembering and reproducing motor acts as an element of the system, receiving and summarizing information from the sensors and closed to the environment through effectors. We describe a hierarchy of levels of the motion memory model and considered in detail the level of trajectory motion control. The results of modeling the functioning of the neural network traffic management.
8. GUNDELAKH F.V., BAKHSHIEV A.V.
1Russian State Scientific Center for Robotics and Technical Cybernetics, Saint-Petersburg
2Peter the Great St. Petersburg Polytechnic University
Modeling the network for coordinated control of elements of the robot's executive system
The paper deals with the problem of controlling the elements of the robot's executive system by using bio-similar control system. It is based on the pulsed neuron model.Multi-level control system, based on neural network, which is able to store and reproduce state of the robot's executive system is proposed, implemented and studied.
9. GABDRAKHMANOVA N.T. , E.S. GAYSIN
1Peoples’ Friendship University of Russia, Moscow
2Ufa State Petroleum Technological University
Neural network model of residual life assessment of vertical steel tanks
The article discusses the problem of estimating the residual life of vertical steel tank. The model identification was made by measuring the technical condition of the tanks. The problem is solved with the use of neural network technology. To obtain a solution confidential intervals were built.
10. BURAKOV M.V.
Saint Petersburg State University of Aerospace Instrumentation
Neural supervisor for tuning the adaptation gain of model reference adaptive control system
The implementation of model reference adaptive control (MRAC) system with tuning the adaptation gain by neural net (NN) subsystem is analyzed in this paper. Proposed feed-forward NN consists of nonlinear first layer and linear output layer. The genetic algorithm is used to search parameters of NN. The results confirm benefits of the proposed controller in comparison with the standard MRAC for DC motor velocity control system.
11. V.N. SHATS
Independent investigator, St. Petersburg
Classification algorithm based on a stochastic granulating of information
In this paper, we proposed a new approach to the problem of classification, based on the allocation in given information of granules classes, objects, feature, and index of feature. For of a classes of objects, established a stochastic dependence between sets of frequency elements of granules and a parameter that characterizes the degree of fragmentation features. The sequence the parameter values determines a sequence for error of decision converging in probability. It was shown that the proposed algorithm has high accuracy, simplicity and universality for any type of data.
12. ЕNGEL E. A.
Katanov Khakass State University, Abakan
The energy saving technology of photovoltaic system’s intelligent control
Within neuro-evolutionary approach, fuzzy logic and Smart Grid the energy saving technology of intellectual control of photovoltaic system organizing the interaction of identification PV system’s mode and control on its basis is developed. The simulation results in Matlab show that the proposed technology achieves competitive performance, as compared to classical control scheme.
13. VS ROSTOVTSEV
Vyatka State University
CHOICE OF METHOD hardware implementation nonlinear transformer NEURON on FPGA
The results of the analysis of the author of a hardware implementation of the sigmoid function of the nonlinear converter based on programmable logic integrated circuits (FPGAs) Cyclone III EP3C120. At high speed requirements for the development of neural network emulators nonlinear transformer at the hardware level is an urgent task. There are ways of implementing a non-linear transducer on the FPGA, their advantages and disadvantages. The materials presented can be useful for developers who develop neural emulators
SESSION 2
Wednesday, April 27 10.00 – 11.30
Lecture-hall Алексеевский зал
Chair: Prof. DOLENKO SERGEY
Applications of neural networks
14. * EFITOROV ALEXANDER OLEGOVICH1Lomonosov Moscow State University
2Skobeltsyn Institute of Nuclear Physics Lomonosov Moscow State University
Binary classification of dynamic system's states by machine learning algorithms
The results of solving the problem of binary classification of the states of the dynamic system by supervised machine learning algorithms. Demonstrated a positive effect of using a sequentially lagged time series values as the input variables to determine the state of the system
15. * POLAVSKA ANNA ANATOLIINVA
State University for Transport Economy and Technologies, Kyiv, Ukraine
BIOLOGICALLY MOTIVATED APPROACH TO PARALLEL-HIERARCHICAL IMAGE PROCESSING
The multistage approach to image processing, described in this paper, comprises main types of cortical multistage convergence. The multistage system highlights spatial regularities, which are passed through a number of transformational levels to generate a coded representation which encapsulates, in a computer manner, structure on different hierarchical levels in the image. At each processing stage a single output result is computed to allow a very quick response from the system.
16. * TEREKHOW V.I., CHERNENKY I.M.
Bauman Moscow State Technical University
INTELLIGENT CONTROL SYSTEM OF PURITY OF THE TERRITORY NEAR THE ENTRANCE OF THE BUILDING
The problem of automating the control of purity of the territory near the entrance of the building is reviewed. The developed control system of purity of the territory near the entrance of the building and the use of neural network approach to analyze the purity of the territory with the help of CCTV cameras in winter is described. The effect of image settings for correct recognition of cleanliness territory is analyzed. The conclusions about the need to create intelligent hybrid control system of purity of the territory near the entrance of the building, based on neural network and expert system to increase system performance is formed.
17. * SHIROKY V.R., DOLENKO S.A., MYAGKOVA I.N., SENTEMOVA N.S.
1Skobeltsyn Institute of Nuclear Physics Lomonosov Moscow State University
2Lomonosov Moscow State University
Study of horizon of forecast of the state of the EARTH magnetosphere with artifiсial neural networks
The paper reviews the results of forecast of the time series of geomagnetic activity indicies Dst and Kp and flux of relativistic electrons with energies >2 MeV with artificial neural networks with horizon from one to twelve hours.The presented forecasts are based on multivariate time series, including the value of the index and electron flux values themselves, as well as data on the parameters of the solar wind and the interplanetary magnetic field during the last day with a resolution of one hour.
18. I.V. ISAEV, S.A. DOLENKO
Skobeltsyn Institute of Nuclear Physics Lomonosov Moscow State University
IMPROVING THE RESISTANCE OF NEURAL NETWORK SOLUTION OF INVERSE PROBLEMS TO NOISE IN DATA BY GROUP DETERMINATION OF PARAMETERS
When solving multi-parameter inverse problems (IP) by neural networks, the problem is usually solved separately for each parameter. In previous studies it was shown that aggregation of parameters into groups with simultaneous determination of the values of all the parameters of the group in some cases allows improving the accuracy of the solution of various multi-parameter IP. In this study, the influence of the observed effect on the stability of neural network solution of an IP against noise in data was investigated.
19. * STAROVEROV BORIS ALEKCANDROVITCH GNATYUK BORISLAV ALEKSEEVITCH
Kostroma State Technological University
CONFIGURING OF NEURAL NETWORK ENSEMBLE AS BASE FOR INVARIANT FORECASTING SYSTEM (EVIDENCE FROM ELECTRICAL ENERGY CONSUMPTION)
Problems of neural network forecasting system, invariant to type of electrical energy consumption schedule are solved. Minimum length input vector structure is explained; neural network ensemble structures are determined; se-lection of the most effective neural network types in the ensemble is held. Original three-level structure of neural network ensemble is developed. Its high forecasting capability makes network perspective for solving information statistical analysis problems.
SESSION 3
Wednesday, April 27 11.30 – 13.00
Lecture-hall Алексеевский зал
Chair: Prof. RED'KO VLADIMIR
Adaptive behavior and evolutionary modelling
20. SOKHOVA Z.B., RED'KO V.G.Scientific Research Institute for System Analysis, Moscow
Agent-based model of agricultural land rental market in а region
A computer model of rent of agricultural land in a small region have been constructed and investigated. Transparent method for leasing agricultural land has been proposed. The model demonstrates the natural dynamics of capitals, resources and capacities of land users.
21. * FEDORENKO YURIY SERGEEVICH, GAPANYUK YURIY EVGENIEVICH
Bauman Moscow State Technical University
THE CONSTRUCTION OF ADAPTIVE MODELS BASED ON MULTILEVEL NEURAL NETWORK USING METAGRAPH APPROACH
The problem of constructing adaptive models to describe the flow of data in real time is considered. The shortcomings of existing algorithms are analyzed. An approach that combines the advantages of deep learning and self-organizing incremental neural networks to build adaptive models is proposed. Metagraph-based approach for describing a software implementation of a neural network is proposed. Implemented part of the proposed algorithm, experiments on clustering data are implemented.
22. CHERNENKIY VALERIY MIKHAILOVICH, TEREKHOV VALERY IGOREVICH, GAPANYUK YURIY EVGENIEVICH
Bauman Moscow State Technical University
REPRESENTATION OF COMPLEX NETWORKS BASED ON METAGRAPHS
The concept of complex networks as "networks with the emergence" is reviewed. The formal model of metagraph is given. The advantages of metagraphs over hypergraphs and hypernetworks to represent "networks with the emergence" are reviewed. The using of holonic multi-agent system for realization of activity in a complex network is proposed. The structure and principles of metagraph agent as an active element for the simulation of complex networks are reviewed.
23. SUSLOV V.V.
Institute of Cytology and Genetics, Siberian Branch of the Russian Academy of Sciences, Novosibirsk
GOAL SETTING AND ORGANISMAL EVOLUTION
Evolutionary controversies that happen as individual organisms explore the novelty, that is, poorly efferent environments, have been considered. The inadequacy of optimization-based scenarios has been demonstrated. Based on the properties of stress (non-specificity; tiredness resulting an interference of biochemical and/or physiological reactions) a non-optimization-based scenario implying goal setting has been proposed.
24. * KOPELIOVICH MIKHAIL VIKTOROVICH, PETRUSHAN MIKHAIL VIKTOROVICH
1Southern Federal University, Rostov-on-Don
2Scientific Research Center of Neurotechnologies, Southern Federal University, Rostov-on-Don
Evolutionary algorithm for structural-parametric optimisation of method of remote photoplethysmography.
The problem of optimal combination of algorithms and parameters of method of remote photoplethysmography for heart rate estimation is considered. To solve this optimization problem evolutionary algorithm that implements adaptive reinforcement learning was developed. Optimal solution was obtained by using original criterium that combines root mean squared error and lifespan of an agent.
25. P.P. DJACHUK (ML), P.P. DJACHUK, S.A. KARABALYK, I.V. SHADRIN
1Siberian Federal University, Krasnoyarsk
2Krasnoyarsk State Pedagogical University named after V. P. Astafyev
Diagnosis of unstable cognitive state of the active agent
The problem of cognitive diagnosis of unstable states of the active agent (AA), the self-consistent problem changing complexity of the environment, defined by the frequency of reinforcements. Experimentally proved the existence of two types AA differing ability to compress the information and change activities as a result of self-regulatory mechanisms of bifurcation. The experiment is interpreted in the model of the brain Hawkins, including hierarchical processing times-tion series of events and allocation of invariants events.
POSTER SESSION 2
Wednesday, April 27 14.00 – 15.00
Lecture-hall Петровский зал
Chair: Prof. CHIZHOV ANTON
Neurobiology
26. DICK O.E.Pavlov Institute of Physiology of the Russian Academy of Sciences, St.Petersburg
Wavelet and multifractal estimation of the intermittent photic stimulation response in the electroencephalogram of patients with cerebrovascular disturbances
The task is to estimate the degree of responses to intermittent photic stimulation for subjects with cerebrovascular disturbances. This degree is evaluated on the basis of analysis of changes in multifractal and wavelet properties of electroencephalographic (EEG) patterns taking place during the photic stimulation. The degree of multifractality of EEG patterns does not change considerably during the stimulation. By contrast, the coefficients of photic driving and holding and the energy increase times gained by the wavelet spectra in EEG patterns of patients with cerebrovascular disturbances differ significantly from the characteristics determined for the healthy subjects.
27. VORONKOV G.S.
Lomonosov Moscow State University
THE PERCEPTION OF SMALL LIGHT SOURCES RIMLESS ALSO CAN MANIFEST IN THE FORM OF THE "SPECIFIC VISION DEFICIT" PATCH
The study of the mysterious "specific vision deficit" (SVD) phenomenon is continued. The essence of the SVD phenomenon is the sensation of seeing a dark patch (SVDp) in a small hole (diaphragm) in the lumen, as previously described. In the work it was found that patch, having all the SVDp characteristics, may also occur in the perception of a number of other light sources. This fact opens new possibilities in the study of the SVDp nature. This paper describes these sources and also the properties of the created by them SVDp.
28. D.A. VOLKOV, O.A. MISHULINA
National Research Nuclear University (MEPhI), Moscow
MECHANISM OF SYNCHRONIZED BURSTING IN THE NEURONAL CULTURE MODEL ON THE BASIS OF NEURON WITH AUTOREGULATION
The paper presents a neural network model that exhibits synchronized bursting activity. It is recurrent network formed on the basis of modified Leaky Integrate-and-Fire neurons. Extended vector of the current state of the neuron is introduced. It contains information about both its own ability to generate spikes and the current activity of the presynaptic neurons. We have refused a widespread model of long-term synaptic plasticity STDP and worked out a new learning rule for synaptic weights. The recurrent neural network with the proposed model of a neuron shows activity in the form of synchronized bursts, as it occurs in neuronal cultures. The simulation results are presented. Prospects of application of the model to study the mechanisms of learning in the neuronal culture in vitro are discussed.
29. A.V. PARASKEVOV, D.K. ZENDRIKOV
1Institute for Information Transmission Problems (Kharkevich Institute), Moscow
2The Moscow Institute of Physics and Technology (State University)
On the Spatial Dynamics of a Network Spike in Neuronal Cultures
It is shown that in the model two-dimensional neuronal cultures there exist a few spatial nucleation centers of a network spike, from where the synchronous spiking activity propagates in the network as single circular traveling waves. The number and spatial locations of the nucleation centers are unique and steady for a given neuronal network, and these quantities vary over the different networks. A functional role of inhibitory neurons is specified qualitatively. The obtained results are consistent with experimental observations.
30. VITYAEV E.E., DERGILEV A.I., CHADAEVA I.V., VASKIN Y.Y., SPITSINA A.M., KULAKOVA E.V., VISHNEVSKY O.V., ORLOV Y.L.
1Sobolev Institute of Mathematics, Siberian Branch of the Russian Academy of Sciences
2Novosibirsk State University
3Institute of Cytology and Genetics, Siberian Branch of the Russian Academy of Sciences, Novosibirsk
Search for regulatory context signals in genomic DNA
Computer search of combinatorial regularities in location of transcription factors binding sites in a genome is important for the study of gene expression regulation in eukaryotes. It challenges new bioinformatics tasks and demand development of specialized software and tools based on intellectual data analysis (Data Mining). It is necessary to integrate the computer tools related to the logical rules techniques, such as packages UGENE and the ExpertDiscovery system.
31. SMIRNITSKAYA IRINA ARKADIEVNA
Scientific Research Institute for System Analysis, Moscow
The role of diferent amygdala nuclei in appetitive and aversive Pavlovian-to-instrumental transfer.
The review focuses on the role of amygdala nuclei in different formes of appetitive and aversive Pavlovian and instrumental behaviors. The neural circuit wich can provide those behaviors is discussed
SESSION 4
Wednesday, April 27 15.30 – 18.00
Lecture-hall Алексеевский зал
Chair: Prof. CHIZHOV ANTON
Neurobiology
32. BRAZHE A.R.Lomonosov Moscow State University
Algorithms to generate realistic astrocyte networks for neuro-glial modeling
Astrocytes have intricate structure and complex dynamics. They play a pivotal role in organizing patterns of neuronal network activity. Theoretical studies of neuron-glia interactions require a way to generate spatial structures resembling real astrocytic networks. This study offers two approaches to create such patterns: a data-driven one and generative approach. Discussed are the ways to account for neuronal synaptic activity in the context of the dynamics of the spatially detailed models of astrocytic networks.
33. BABENKO V N, BRAGIN AO, MEDVEDEVA IV, CHADAEVA IV, DERGILEV AI, SPITSYINA AM, KUDRIAVTSEVA NN, MARKEL AL, ORLOV YUL
1Institute of Cytology and Genetics, Siberian Branch of the Russian Academy of Sciences, Novosibirsk
2Sechenov University
Analysis of aggressive behaviour by whole RNA sequencing in brain compartments
The development of molecular techniques in recent years has opened up the prospects for engagement in the study of gene regulation mechanisms in complex behaviors, including aggressive behavior. On the modeling rat species we present an analysis of transcriptomic data (RNA-seq) of the brain compartments (hypothalamus, ventral tegmental area, midbrain raphe nuclei). We found genes with increased expression in the studied regions of the brain presumably associated with behavior.
34. CHIZHOV A.V., VERHLUTOV V.M., SMIRNOVA E.YU., AMAKHIN D.V., ZAITSEV A.V.
1Ioffe Physical Technical Institute, Russian Academy of Sciences, St Petersburg
2Institute of Higher Nervous Activity and Neurophysiology of RAS, Moscow
3Sechenov Institute of Evolutionary Physiology and Biochemistry of the Russian Academy of Sciences, St.Petersburg
Wave-like propagation of preictal discharges
Preictal discharges are observed in epileptiform neural tissue as travelling waves of hypersynchronious neuronal activity. Their generation and spread are studied with the help of biophysically detailed model of a neural tissue as a spatially distributed continuum of excitatory recurrently connected populations of adaptive Hodgkin-Huxley-like neurons, based on the conductance-based refractory density approach. Also, the model demonstrates that the preictal discharge propagation can not be prevented by a localized cut of connections.
35. SMIRNOVA ELENA YURIEVNA, CHIZHOV ANTON VADIMOVICH
Ioffe Physical Technical Institute, Russian Academy of Sciences, St Petersburg
The minimal model of the color and orientation processing in the primary visual cortex
We propose a minimal model of color and orientation processing in the primary visual cortex, based on novel experimental data. The model is constructured as a network of neuronal populations distributed in the feature space of color and orientation. The network has a toroidal topology. The populations are described by the rate model. The torus-model reproduces known characteristics of the color-sensitive neurons’ behavior and predicts some effects associated with a perception of color hue for gray non-oriented or two-color stimuli.
36. N.G. BIBIKOV
N.N. Andreyev Acoustics Institute, Moscow
Temporal analysis of the sound in auditory system of the frog
We investigated the mechanisms of temporal analysis of audio signals in the frog’s medulla and midbrain neurons. Extracellular activity was recorded in two nuclei of the auditory pathway of the grass frog (Rana t. temporary): dorsal medullar nucleus and torus semicircularis. An analysis of the statistical properties of background activity provided information about intrinsic properties of single units. The slow changes in unit’s sensitivity demonstrates the adaptive behavior of auditory neuronal ensembles. In the process of neuron’s adaptation to continuous signal, its ability to recognize small amplitude changes improves significantly. We studied signal envelope features that caused neuronal spike generation during long tone that was amplitude modulated by repeatable -wave of envelope function, but more central neurons were often responsive to the rate of change in amplitude and even to its acceleration.
37. PARIN S.B., POLEVAYA S.A., KOVALSHUK A.V., GROMOV K.N., CHERNOVA M.A.,YACHNO V.G.
1N.I. Lobachevsky State University of Nizhni Novgorod
2The Institute of Applied Physics of the Russian Academy of Sciences, Nizhny Novgorod
By constructing a model management heart rate variability stress
The results of mathematical modeling and experimental system of governance heart rate variability during stress. Checked the hypothesis that the beginning of the first stage of the stress caused by the activation of neuronal (catecholamines as neurotransmitters) and hormonal components sympathoadrenal and neuronal (enkephalins) component of the endogenous opioid system. We discuss the constraints and growth point of the proposed model.
38. DERGILEV A.I., SVICHKAREV A.V., CHADAEVA I.V., ABNIZOVA I.I., KULAKOVA E.V., SUBKHANKULOVA T.N., SUSLOV V.V., NAUMENKO F.M., VITYAEV E.E., ORLOV Y.L.
1Novosibirsk State University
2Peter the Great St. Petersburg Polytechnic University
3Institute of Cytology and Genetics, Siberian Branch of the Russian Academy of Sciences, Novosibirsk
4Sechenov University
Computer analysis of transcription factor binding sites clusters in genome
The analysis of molecular mechanisms of gene expression regulation challenges new methods of genome data analysis. The growing volume of sequencing data including experimentally defined transcription factor binding sites makes more complex the problem of regulatory signal analysis on nucleotide sequences. By informatics methods we considered ChIP-seq data clusters of binding sites in mouse genome.
39. T.A. ZAPARA, A.L. PROSKURA, S.O. VECHKAPOVA, A.S. RATUSHNYAK
1
2The Institute of Computational Technologies of SB RAS (ICT SB RAS), Novosibirsk
MOLECULAR MECHANISMS OF CHANGE EFFICIENCY OF SYNAPTIC TRANSMISSION IN EXCITATORY SYNAPSES
The molecular organization and temporal dynamics of change in the efficiency of interneuronal communication is investigated. Integrative activity of molecular systems spiny synapses of pyramidal neurons of hippocampal Schaffer collaterals system, ensuring the restructuring of synaptic connections based on incorporation/removal from membranes of various types of glutamate receptors is discussed.
POSTER SESSION 3
Thursday, April 28 14.00 – 15.00
Lecture-hall Петровский зал
Chair: Prof. RED'KO VLADIMIR
Adaptive behavior and evolutionary modelling
40. V.G.REDKO, T.I. SHARIPOVA, G.A. BESKHLEBNOVAScientific Research Institute for System Analysis, Moscow
THE USE OF NEURAL GAS METHOD FOR THE SIMULATION OF THE SEARCH BEHAVIOR OF AGENTS
Method time growing neural gas model studies the behavior of the organism (agent), which need power, security, knowledge. The results of modeling agent in one-dimensional and two-dimensional labyrinths. The possibility of a sharp reduction in the size of neural network in which it is enough to remember only those points of space, which varies in the surrounding situation.
41. LAVROV V.V., RUDINSKY A.V.
1
2Center of System Consulting And Education "Synergia", St.Petersburg
FORMATION OF THE IMAGE AS THE MATRIX OF INFORMATION FRAGMENTS OF VARYING WEIGHT
The images that are stored in memory as model representatives of real objects are assembled, first, inadvertently and, secondly, consciously and purposefully. The image is built in the form of a matrix under the control of motivation, which is due to urgent need, define the behaviors. In fact, the transformation of the image is similar to training when under the influence of the experience changing the weight information of the fragments and rebuild the subsystem types. This opens the prospect of studies of the properties of the image in connection with aggression.
POSTER SESSION 4
Thursday, April 28 14.15 – 15.00
Lecture-hall Петровский зал
Chair: Prof. LITINSKII LEONID
Neural network theory
42. CHEREDNIKOV D.Y.; SHIBZUKHOV Z.M.Moscow State Pedagogical University
About sigma-pi neurons of aggregation type
A new class of models of artificial neurons is described in this work. These models are based on the following principles: 1) contributions of synapses are summed with the help of certain aggregation operation; and 2) contribution of complex synapse or synaptic cluster is computed with the help of another aggregation operation on the set of simple synapses. These models include a big part of the known functional models of neurons. For a class of the aggregating neurons generalizing model sigma-pi neuron, it is shown that they can be correctly trained on the final sets of precedents.
43. SMOLIN V.S.
Keldysh Institute of Applied Mathematics, Moscow
The outlook of self-organizing maps applicability for solving complex applied tasks
All neuroinformational (NI) models perform input signals X transformations into output signals Y. The NI model connections weights matrix after adaptation defines the tabular function Y=F(X,A), where A is NI model elements activity. For several reasons very complex functions couldn’t be realized by one table. The main reason is “dimensionality damnation”, interpolation basic points number exponential growth with Y=F(X,A) dimensionality increase. Some ways of very complex functions approximation by NI models set are suggested.
44. BONDAREV V.N.
Sevastopol State University
TRAINING RULES OF PULSE NEURON FOR THE ADAPTIVE SIGNAL PROCESSING
The generalized models of a multi-input pulse neural element in a vector-matrix form focused on the solution of digital signal processing problems are offered. The adaptive filtration scheme based on pulse neuron is considered and rules of its training for signal restoration and noise suppression and for blind deconvolution are developed. Results of computer simulation are given.
POSTER SESSION 5
Thursday, April 28 14.15 – 15.00
Lecture-hall Петровский зал
Chair: Prof. KAZANOVICH YAKOV
Neural networks and cognitive sciences
45. CHERNYSHEV ALEXEY SERGEEVICHBauman Moscow State Technical University
Bayesian optimization of spiking neural nets for time series classification task
This work contains the application of spiking neural networks to time series classification task. Because of the lack of mathematical framework for such biologically inspired neural networks, this work tries to solve optimization task of parameters of the network with surrogate models. Quality metric for classification performance designed.
46. CHUGROVA M., BAKHCHINA A., PARIN S.
1N.I. Lobachevsky State University of Nizhni Novgorod
2Institute of Psychology of Russian Academy of Sciences, Moscow
Investigation of dynamics of functional state of participants in a group discussion
We research the problem of the interaction of the functional state of human and social environment in which he is located. We have studied a form of social interaction in a polylogue communication in a group discussion, its impact on functional hemispheric asymmetry the participants and the dynamics of the parameters of heart rate variability.It was shown that changes in the functional state in the process of group discussion related to the individual psychological characteristics of its participants.
47. BAKSHEEVA Y., DANILOVA I., SAPOZHNIKOVA K., TAYMANOV R.
1Saint Petersburg State University of Aerospace Instrumentation
2D.I.Mendeleyev Institute for Metrology
Metrological approach to substantiation of a non-linear character of emotion formation
The paper deals with special features of the metrological approach to measurements of emotions. The results are given, which prove the validity of a hypothesis related to the nonlinear conversion of musical and other acoustic impacts into neurophysiologic reactions. These results also show that functions of basic blocks of a measurement model, which links musical impacts with the emotions of listeners, are valid.
48. TAYMANOV R., KOSTROMINA S., SAPOZHNIKOVA K.
1D.I.Mendeleyev Institute for Metrology
2Sanct-Petersburg State University
Formation of a scale of emotions taking place under the influence of music
A critical analysis of emotion scales that are applied while evaluating the perception of music and its fragments, is given. The difference of requirements is shown that is characteristic for psychologists and musicologists, on one hand, and metrologists developing the instrument assigned for measurements of expected listener’s emotions, on the other hand. The scale acceptable for the appropriate measurement model should rely on the concept of “emotion language”. A version of such scale is justified.
49. ALEXANDR Y. PETUKHOV, SOFIA A. POLEVAYA
1
2N.I. Lobachevsky State University of Nizhni Novgorod
Modeling of communicative individual interactions through the theory of information images
The current work represents a formalized description of information and communicative interactions of individuals on the basis of theory of information images. It also demonstrates how important it is to choose the models type adequate to the systems under research. It also introduces an explication of the possibility to create the model of information and communication interactions that can illustrate transmission of information between two and more individuals. There is also an example of modeling on the basis of the current theory taking into account the results of the experiment (bilingual Stroop test).
50. KRYLOV A.K.
Institute of Psychology of Russian Academy of Sciences, Moscow
Modeling of improving a skill and the interference effect
A model of formation and reorganization of individual experience is proposed based on systemic psychophysiology. The model applied for an investigation of learning speed depending on the sequence of the learning tasks. The effects of improving a skill and interference are shown in the computer experiments with the model.
51. MEILIKHOV E.Z., FARZETDINOVA R.M.
National Research Centre "Kurchatov Institute", Moscow
Multiple goal pursuit -- to kill two birds with one stone or to fall between two stools ?
We present the simple phenomenological (but - analytic) model allowing to formalize description of multitasking, i.e. simultaneous performing several tasks. That process requires distribution of attention, and for great number of goals do not lead to success. Our consideration shows that simultaneous performing more than two tasks is, most likely, impossible.
52. KHARYBINA Z.S.
Southern Federal University, Rostov-on-Don
Construction of entorhinal grid cell firing maps with distinct grid scales and orientations
The problem under research is firing modelling of grid cells from distinct scale and orientation modules with the use of even cyclic inhibitory networks. The algorithm of transformation of informational unit otputs into grid cell firing has been developed. Grid cell firing maps with distinct grid scales and orintations have been constructed.
53. RATUSHNYAK ALEXANDER, PROSKURA ANNA, ZAPARA TATYANA, SOROKOUMOV EVGENIJ
1
2The Institute of Computational Technologies of SB RAS (ICT SB RAS), Novosibirsk
PRINCIPLES AND MOLECULAR MECHANISMS OF FUNCTIONING NERVE CELLS
To understand the principles and mechanisms of the brain requires knowledge not only of the interactions of nerve cells and structures, but, above all, an understanding of the functional properties of neurons and the molecular organization of these basic elements allowing neural networks to solve the information problems. Simplification of nerve cell model is the main cause of the lack of significant results and for their receiving necessary to integrate existing knowledge, develop a conceptual model of functional architecture of the neuron.
54. TARASOV D.S.
Meanotek, Kazan
Preserving personal conversational style and diversity in neural conversational models
Neural conversational models are subject of active research interest. Trained on millions of dialog turns, they are capable of simulating conversation with humans and provide assistance to users. However, after training on noise corpus of conversations between many users, these models exhibits inconsistent personal traits and conversational styles. This paper presents a method to apply specific personal conversation traits to neural dialog models, trained on large heterogeneous dialog corpus such as OpenSubtitles.
55. MEILIKHOV E.Z., FARZETDINOVA R.M.
National Research Centre "Kurchatov Institute", Moscow
ON MILLER’S RULE FOR THE BRAIN WORKING MEMORY, OR WHY HUMAN MEMORY IS SO SHORT
Working memory is a cognitive construct that describes how infor-mation can be maintained in brain for a limited period of time, while concurrent processing is also performed. We present a simple model that accounts for working memory span and explains the origin of the cognitive Miller's rule (Magical Number Seven).
SESSION 5
Thursday, April 28 15.30 – 18.00
Lecture-hall Алексеевский зал
Chair: Prof. LITINSKII LEONID
Neural network theory
56. KISELEV M.V.The Chuvash state university named after I. N. Ulyanov
Asynchronous/polychronous scheme of information coding in spiking neural networks
A novel approach to information coding in spiking neural networks is proposed. It is based on asynchronous activity of ensembles of polychronous neuronal groups - sets of neurons firing with the strictly fixed time delays one after one. These ensembles are formed in the considered neural networks due to STDP-based synaptic plasticity mechanism.
57. M. E. MAZUROV
Plekhanov Russian University of Economics
MATHEMATICAL MODEL OF THE DYNAMIC PULSE THE BLOCKING OF THE NEURON
Mathematical models of dynamic pulse or spike of neurons, including synaptic system and excitable system, generating individual pulses or pulse sequences – Barstow. A mathematical model of the blocking of the neuron, based on a blocking oscillator having a number of remarkable properties: the ability to generate pulses that are close in the form of neural impulses, bursty and much more. The effectiveness of the blocking of neuron confirmed in the computational experiment.
58. GAI V. E., UTROBIN V. A., GAI N. V.
Nizhny Novgorod State Technical University named after R.E. Alekseev
Modeling of associative memory from the point of view of theory of active perception
We consider the problem of creating a model of associative memory based on the theory of active perception. It is proposed to use this model for solving the problem of the search for similar images. We reduce the resultsof computational experiments comprising the temporary assessment of the proposed models.
59. E.S.GRICHUK, M.G.KUZMINA, E.A.MANYKIN
1National Research Centre "Kurchatov Institute", Moscow
2Keldysh Institute of Applied Mathematics, Moscow
3National Research Nuclear University (MEPhI), Moscow
A model of pulsed oscillator network and perspectives of its application to problems of information routing in wireless sensor networks
Development of performance principles of wireless sensor networks (WSN) is currently of special interest. An oscillatory network model with pulsed oscillator dynamics and pulsed oscillator interaction is suggested in the paper. Design of appropriate network connectivity rules, capable to guarantee controllable network synchronization, is discussed.
60. VASILYEV A.N., LAZOVSKAYA T.V., TAKHOV D.A., SHEMYAKINA T.A.
Peter the Great St. Petersburg Polytechnic University
Neural network approach to solving complicated problems for ordinary differential equations
Three model problems for ordinary differential equations whose conventional methods solving encounters serious difficulties of different kinds are considered. It is shown how these problems can be solved by a single neural network approach. Similarly it is possible to overcome difficulty of other types.
61. KOZLOV D.S., TIUMENTSEV YU.V.
Moscow Aviation Institute (National Research University)
Neural network based semi-empirical models for dynamical systems represented by differential-algebraic equations of index-1
A simulation problem is discussed for nonlinear controlled dynamical systems represented by differential-algebraic equations of index 1. It is proposed to seek a solution of the problem within the semi-empirical modeling approach. The results of simulation for elaboration procedure of lift coefficient in respect to reentry hypersonic vehicle are presented. Equilibrium glide part of zero-thrust phase of a flight is described.
62. VLASOV D.S., SBOEV A.G., SERENKO A.V., RYBKA R.B., MOLOSHNIKOV I.A.
1National Research Centre "Kurchatov Institute", Moscow
2The Moscow Institute of Physics and Technology (State University)
To the possibility of a spiking neuron to learn the task of binary vectors classification by long-term synaptic plasticity STDP
A study of possibility of the Leaky Integrate-and-Fire neuron model to learn to classify inputs encoding binary vectors with a supervised protocol based on long-term synaptic plasticity STDP is presented. The synaptic weights values with which the neuron does the task of binary vectors classification are shown to be able to emerge as the result of learning. The STDP model parameters with which non-bimodal weights distributions can be reached during learning are found.
63. L. YU. ZHILYAKOVA, O. P. KUZNETSOV
V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Moscow
Principles of discrete simulation heterogeneously mechanisms in nervous systems
New principles for the functioning of neurons are proposed. The model of neurobiological neuron formalizes the diverse types of neurons and is an extension of the model by McCulloch and Pitts. Neuron, analogously to the classic ("electrical") model, can be in one of two states (active and passive); however, unlike electric models, it "sees" not all the signals that it receives, but only those, which correspond to the types of its receptors.
64. M.S. TARKOV
Rzhanov Institute of Semiconductor Physics, Siberian Branch of Russian Academy of Sciences, Novosibirsk
HOPFIELD NETWORK WITH INTERNEURONAL CONNECTIONS BASED ON MEMRISTOR BRIDGES
A scheme is proposed for the Hopfield network’s hardware implementation with interneuronal connections through bridges using memristors. The method of calculating the bridge memristor resistance values by weighting coefficients is developed. It is shown how to use the CMOS transistor switches to control the memristance (memristor resistance) value.
65. BARANOV D.S., BOGOMOLOV Y.V.
P.G. Demidov Yaroslavl State University
Chaotic dynamics and multistability phenomenon in one neural network model
Features of the dynamics of one neural network model, based on three McCulloch-Pitts formal neurons are investigated in the submitted clause. For this neural network model there are summarized older results which character-ized different behavior features (including chaotic dymanics), and the scale invariance and multistability phenomenon are also registered.
SESSION 6
Friday, April 29 12.00 – 13.00
Lecture-hall Алексеевский зал
Chair: Prof. TEREKHOV SERGE
Neural networks and cognitive sciences
66. S.V. BOZHOKIN, I.B. SUSLOVAPeter the Great St. Petersburg Polytechnic University
INSTANTANEOUS WAVELET PHASE CORRELATION OF ELECTROEHCEPHALOGRAM SIGNALS
We developed a new algorithm to detect instantaneous phase correlation of nonstationary electroencephalogram (EEG) signals registered from different leads of the cerebral cortex. The method is based on the continuous wavelet transform (CWT), using the Morlet mother wavelet with a control parameter that allows you to change the spectral and temporal resolution of the signals under study. The algorithm includes phase straightening procedure, eliminating the critical points in the CWT phase behavior at the boundaries of the interval [0;2Pi]. We tested the approach by calculating phase correlation in time of two EEG signals.
67. N.N. DANILOVA, S.I. SEMENYUK, E.A. STRABYKINA
Lomonosov Moscow State University
Interhemispheric difference of the C. Mangina test in the group of young people, specialized in mathematics
Mathematical abilities and reading/ reading comprehension abilities of the C. Mangina Test were investigated by the method of “Microstructural analysis of oscillatory brain activity”. The left hemisphere plays a key role in solving “mathematical” tasks with the increasing of frequency-selective theta and alpha generators activity. Solving tasks on “reading” is implemented by the right hemisphere with increasing of the theta generators activity and with reduction of the alpha generators. The function of the local inhibition is discussed.
68. YURI SKOBTCOV, SAMER EL-KHATIB
1Saint Petersburg State University of Aerospace Instrumentation
2Southern Federal University, Rostov-on-Don
Medical image segmentation using mixed and multi-elitist exponential particle swarm optimization
Proposed two image segmentation algorithms: mixed k-means and particle swarm optimization and multi-elitist exponential particle swarm optimization. Developed system for image segmentation using proposed above algorithms. Done testing of proposed algorithms using Berkley benchmark images with different initial conditions, done results comparison with another image segmentation algorithms. Done segmentation precision comparison with another existing algorithms.
69. A.D. CHERKAY
1Moscow Aviation Institute (National Research University)
2Novosibirsk State University
LANGUAGE OF HEART RHYTHM
The article discusses proposed by the authors in the 70s of XX century, the method of allocating in the generated pulse signals repeatable of sequences an organism belonging to e-tube with a given e, which provides identification of the registered series RR intervals ECG repeatable fragments - "words" ("patterns", " matrices ") of the language of the heart rhythm.
SESSION 7
Friday, April 29 14.00 – 15.30
Lecture-hall Алексеевский зал
Chair: Prof. KAZANOVICH YAKOV
Applications of neural networks
70. S.V. BEZOBRAZOV, J.M. BLACKLEDGE1Brest State Technical University, Belarus
2University of KwaZulu-Natal
PERSONAL CRYPTOGRAPHY USING ARTIFICIAL NEURAL NETWORKS
The paper presents a method of generating encryption algorithms using neural networks. Based on the application of natural noise sources we ‘teach’ a system to approximate the input noise with the aim of generating an output nonlinear sequence. This output is then treated as an iterator which is subjected to a range of tests to check for potential cryptographic strength in terms of metric. This approach provides the potential for generating an unlimited number of unique PRNG.
71. DUDKIN ALEXANDER, MARUSHKO EVGENY
United Institute of Informatics Problems of the National Academy of Sciences of Belarus, Minsk
STUDY OF THE EFFECTIVENESS OF ENSEMBLES OF NEURAL NETWORKS FOR FORECASTING TIME SERIES TELEMETRY SPACECRAFT
We consider the problem of forecasting multivariate time series telemetry data of the spacecraft. The analysis of the approaches to build ensembles of neural networks from the viewpoint of prediction accuracy. We investigate the possibility of teaching ensembles of neural networks.
72. L.A. STANKEVICH, K.M. SONKIN, N.B. SHEMYAKINA, ZH.V. NAGORNOVA, J.G. KHOMENKO, D.S. PERETS, A.V. KOVAL
1Peter the Great St. Petersburg Polytechnic University
2Sechenov Institute of Evolutionary Physiology and Biochemistry of the Russian Academy of Sciences, St.Petersburg
3International Scientific Center "Arktika", FEB RAS, Magadan
4Bekhtereva Human Brain Institute RAS, St.Petersburg
Development of real time brain computer interface on the basis of neurological classifiers committee of EEG signal
The study is dedicated to the development of real time brain-computer interface (BCI) on the basis of neurological classifier of EEG patterns associated with the imaginary movements of fingers of the one hand. It was shown that such task can be solved using neurofeedback and minimal system time delays. The realized variant of real time BCI with the use of neurological committee of classifiers on the basis of artificial neural networks and support vector machine is presented.
73. FILATOVA N.N., SIDOROV K.V., KHANEEV D.M.
Tver State Technical University
Classification of the human emotions sign by EEG and speech signals at audio-visual or olfactory stimulation
We consider the biotechnical system for the classification a sign of human emotions by analyzing EEG and speech signals by using the features that characterize changes in the properties of attractors reconstructed from samples of these signals. The method of recording speech samples and EEG is proposed. We describe the model of formalized description of speech signals and rules for their classification in accordance with the sign of the emotions experienced by man. Presents the results of a comparison of the findings with the results of analogues.
74. SUKONKIN ILYA NIKOLAEVICH
National Research Nuclear University (MEPhI), Moscow
Training of RBF networks based on the SV-criterion for data clustering problem
A method for training neural networks with radial basis neurons for data clustering problem is proposed. Training is carried out by an evolutionary algorithm with the criterion based on standard volume of clusters(SV-criterion). The proposed neural network method allows to deal with some problems that occur when setting up the conventional models of elliptical clusters. Advantages of the method are demonstrated on simulated and real-world examples.
75. TARASOV D.S.
Meanotek, Kazan
Neural network model for general domain question answering
Question answering is one of long-standing and most important problems of natural language processing. Practical question answering systems are usually based on using of complex linguistic features and problem-oriented knowledge bases. In this work we propose novel neural network model, capable of answering questions without topic restriction by reading natural language documents provided by simple information retrieval methods.
76. KARLOV I.A., KOSHUR V.D.
1National Research University "Higher School of Economics", Moscow
2Siberian Federal University, Krasnoyarsk
ANALYSIS THE RESTORATION OF IMAGES BASED ON NEURAL NETWORK AND CLASSICAL METHODS OF DATA RECOVERY
This paper considers the problem of restoring missing fragments in the images. A brief overview of existing approaches for image restoration. The analysis and testing of individual methods for image restoration for a number of images to assess the quality of reconstruction of images and the possibility of using hybrid models for data recovery.
