DSpace Collection: Volume 14 Number 3
http://hdl.handle.net/10525/648
The Collection's search engineSearch the Channelsearch
http://sci-gems.math.bas.bg/jspui/simple-search
Searching for Nearest Strings with Neural-Like String Embedding
http://hdl.handle.net/10525/695
Title: Searching for Nearest Strings with Neural-Like String Embedding<br/><br/>Authors: Sokolov, Artem<br/><br/>Abstract: We analyze an approach to a similarity preserving coding of symbol sequences based on neuraldistributed representations and show that it can be viewed as a metric embedding process.Selecting Classifiers Techniques for Outcome Prediction Using Neural Networks Approach
http://hdl.handle.net/10525/694
Title: Selecting Classifiers Techniques for Outcome Prediction Using Neural Networks Approach<br/><br/>Authors: Shatovskaya, Tatiana<br/><br/>Abstract: This paper presents an analysis of different techniques that is designed to aid a researcher indetermining which of the classification techniques would be most appropriate to choose the ridge, robust andlinear regression methods for predicting outcomes for specific quasi-stationary process.Double-Wavelet Neuron Based on Analytical Activation Functions
http://hdl.handle.net/10525/693
Title: Double-Wavelet Neuron Based on Analytical Activation Functions<br/><br/>Authors: Bodyanskiy, Yevgeniy; Lamonova, Nataliya; Vynokurova, Olena<br/><br/>Abstract: In this paper a new double-wavelet neuron architecture obtained by modification of standard waveletneuron, and its learning algorithm are proposed. The offered architecture allows to improve the approximationproperties of wavelet neuron. Double-wavelet neuron and its learning algorithm are examined for forecasting non-stationary chaotic time series.Growing Neural Networks Using Nonconventional Activation Functions
http://hdl.handle.net/10525/692
Title: Growing Neural Networks Using Nonconventional Activation Functions<br/><br/>Authors: Bodyanskiy, Yevgeniy; Pliss, Iryna; Slipchenko, Oleksandr<br/><br/>Abstract: In the paper, an ontogenic artificial neural network (ANNs) is proposed. The network uses orthogonalactivation functions that allow significant reducing of computational complexity. Another advantage is numericalstability, because the system of activation functions is linearly independent by definition. A learning procedure forproposed ANN with guaranteed convergence to the global minimum of error function in the parameter space isdeveloped. An algorithm for structure network structure adaptation is proposed. The algorithm allows adding ordeleting a node in real-time without retraining of the network. Simulation results confirm the efficiency of theproposed approach.Linear Classifiers Based on Binary Distributed Representations
http://hdl.handle.net/10525/691
Title: Linear Classifiers Based on Binary Distributed Representations<br/><br/>Authors: Rachkovskij, Dmitri<br/><br/>Abstract: Binary distributed representations of vector data (numerical, textual, visual) are investigated inclassification tasks. A comparative analysis of results for various methods and tasks using artificial and real-world data is given.Neural Network Based Optimal Control with Constraints
http://hdl.handle.net/10525/690
Title: Neural Network Based Optimal Control with Constraints<br/><br/>Authors: Toshkova, Daniela; Toshkov, Georgi; Kovacheva, Todorka<br/><br/>Abstract: In the present paper the problems of the optimal control of systems when constraints are imposed onthe control is considered. The optimality conditions are given in the form of Pontryagin’s maximum principle. Theobtained piecewise linear function is approximated by using feedforward neural network. A numerical example is given.Generalizing of Neural Nets: Functional Nets of Special Type
http://hdl.handle.net/10525/689
Title: Generalizing of Neural Nets: Functional Nets of Special Type<br/><br/>Authors: Donchenko, Volodymyr; Kirichenko, Mykola; Krivonos, Yuriy<br/><br/>Abstract: Special generalizing for the artificial neural nets: so called RFT – FN – is under discussion in the report.Such refinement touch upon the constituent elements for the conception of artificial neural network, namely, thechoice of main primary functional elements in the net, the way to connect them(topology) and the structure of thenet as a whole. As to the last, the structure of the functional net proposed is determined dynamically just in theconstructing the net by itself by the special recurrent procedure. The number of newly joining primary functionalelements, the topology of its connecting and tuning of the primary elements is the content of the each recurrentstep. The procedure is terminated under fulfilling “natural” criteria relating residuals for example. The functionalproposed can be used in solving the approximation problem for the functions, represented by its observations, forclassifying and clustering, pattern recognition, etc. Recurrent procedure provide for the versatile optimizingpossibilities: as on the each step of the procedure and wholly: by the choice of the newly joining elements,topology, by the affine transformations if input and intermediate coordinate as well as by its nonlinear coordinatewise transformations. All considerations are essentially based, constructively and evidently represented by themeans of the Generalized Inverse.Automated Problem Domain Cognition Process in Information Systems Design
http://hdl.handle.net/10525/688
Title: Automated Problem Domain Cognition Process in Information Systems Design<br/><br/>Authors: Loginov, Maxim; Mikov, Alexander<br/><br/>Abstract: An automated cognitive approach for the design of Information Systems is presented. It is supposed tobe used at the very beginning of the design process, between the stages of requirements determination andanalysis, including the stage of analysis. In the context of the approach used either UML or ERD notations maybe used for model representation. The approach provides the opportunity of using natural language textdocuments as a source of knowledge for automated problem domain model generation. It also simplifies theprocess of modelling by assisting the human user during the whole period of working upon the model (using UMLor ERD notations).Compression Technologies of Graphic Information
http://hdl.handle.net/10525/687
Title: Compression Technologies of Graphic Information<br/><br/>Authors: Fesenko, Nikolay<br/><br/>Abstract: The classification of types of information redundancy in symbolic and graphical forms representation ofinformation is done. The general classification of compression technologies for graphical information is presentedas well. The principles of design, tasks and variants for realizations of semantic compression technology ofgraphical information are suggested.Structural Analysis of Contours as the Sequences of the Digital Straight Segments and of the Digital Curve Arcs
http://hdl.handle.net/10525/686
Title: Structural Analysis of Contours as the Sequences of the Digital Straight Segments and of the Digital Curve Arcs<br/><br/>Authors: Kalmykov, Vladimir<br/><br/>Abstract: Recognition of the object contours in the image as sequences of digital straight segments and/or digitalcurve arcs is considered in this article. The definitions of digital straight segments and of digital curve arcs areproposed. The methods and programs to recognize the object contours are represented. The algorithm torecognize the digital straight segments is formulated in terms of the growing pyramidal networks taking intoaccount the conceptual model of memory and identification (Rabinovich [4]).A Partition Metric for Clustering Features Analysis
http://hdl.handle.net/10525/685
Title: A Partition Metric for Clustering Features Analysis<br/><br/>Authors: Kinoshenko, Dmitry; Mashtalir, Vladimir; Shlyakhov, Vladislav<br/><br/>Abstract: A new distance function to compare arbitrary partitions is proposed. Clustering of image collectionsand image segmentation give objects to be matched. Offered metric intends for combination of visual featuresand metadata analysis to solve a semantic gap between low-level visual features and high-level human concept.Decision Trees for Applicability of Evolution Rules in Transition P Systems
http://hdl.handle.net/10525/684
Title: Decision Trees for Applicability of Evolution Rules in Transition P Systems<br/><br/>Authors: Fernandez, Luis; Arroyo, Fernando; Garcia, Ivan; Bravo, Gines<br/><br/>Abstract: Transition P Systems are a parallel and distributed computational model based on the notion of thecellular membrane structure. Each membrane determines a region that encloses a multiset of objects andevolution rules. Transition P Systems evolve through transitions between two consecutive configurations that aredetermined by the membrane structure and multisets present inside membranes. Moreover, transitions betweentwo consecutive configurations are provided by an exhaustive non-deterministic and parallel application of activeevolution rules subset inside each membrane of the P system. But, to establish the active evolution rules subset,it is required the previous calculation of useful and applicable rules. Hence, computation of applicable evolutionrules subset is critical for the whole evolution process efficiency, because it is performed in parallel inside eachmembrane in every evolution step. The work presented here shows advantages of incorporating decision trees inthe evolution rules applicability algorithm. In order to it, necessary formalizations will be presented to consider thisas a classification problem, the method to obtain the necessary decision tree automatically generated and thenew algorithm for applicability based on it.Applications of Radial Basis Neural Networks for Area Forest
http://hdl.handle.net/10525/683
Title: Applications of Radial Basis Neural Networks for Area Forest<br/><br/>Authors: Castellanos, Angel; Martinez Blanco, Ana; Palencia, Valentin<br/><br/>Abstract: This paper proposes a new method using radial basis neural networks in order to find the classificationand the recognition of trees species for forest inventories. This method computes the wood volume using a set ofdata easily obtained. The results that are obtained improve the used classic and statistical models.DNA Simulation of Genetic Algorithms: Fitness Computation
http://hdl.handle.net/10525/682
Title: DNA Simulation of Genetic Algorithms: Fitness Computation<br/><br/>Authors: Calvino, Maria; Gomez, Nuria; Mingo, Luis<br/><br/>Abstract: In this paper a computational mode is presented base on DNA molecules. This model incorporates thetheoretical simulation of the principal operations in genetic algorithms. It defines the way of coding of individuals,crossing and the introduction of the individuals so created into the population. It resolves satisfactorily theproblems of fitness coding. It shows also the model projection for the resolution of TSP. This is the basic step thatwill allow the resolution of larger examples of search problems beyond the scope of exact exponentially sizedDNA algorithms like the proposed by Adleman [Adleman, 1994] and Lipton [Lipton, 1995].<br/><br/>Description: * This work has been partially supported by Spanish Project TIC2003-9319-c03-03 “Neural Networks andNetworks of Evolutionary Processors”.Logic Based Pattern Recognition - Ontology Content (1)
http://hdl.handle.net/10525/681
Title: Logic Based Pattern Recognition - Ontology Content (1)<br/><br/>Authors: Aslanyan, Levon; Castellanos, Juan<br/><br/>Abstract: Pattern recognition (classification) algorithmic models and related structures were considered anddiscussed since 70s: – one, which is formally related to the similarity treatment and so - to the discreteisoperimetric property, and the second, - logic based and introduced in terms of Reduced Disjunctive NormalForms of Boolean Functions. A series of properties of structures appearing in Logical Models are listed andinterpreted. This brings new knowledge on formalisms and ontology when a logic based hypothesis is the modelbase for Pattern Recognition (classification).<br/><br/>Description: * The research is supported partly by INTAS: 04-77-7173 project, http://www.intas.beOn Logical Correction of Neural Network Algorithms for Pattern Recognition
http://hdl.handle.net/10525/680
Title: On Logical Correction of Neural Network Algorithms for Pattern Recognition<br/><br/>Authors: Aslanyan, Levon; de Mingo, Luis; Castellanos, Juan; Ryazanov, Vladimir; Chelnokov, Fedor; Dokukin, Alexander<br/><br/>Abstract: The paper is devoted to the description of hybrid pattern recognition method developed by researchgroups from Russia, Armenia and Spain. The method is based upon logical correction over the set ofconventional neural networks. Output matrices of neural networks are processed according to the potentialityprinciple which allows increasing of recognition reliability.