DOI: 10.7256/2306-4196.2013.1.8314

Abstract: The article presents the components and header modules written in Object Pascal, allowing to easy include libraries with such functions and procedures for numerical solution of linear-algebra as BLAS, LAPACK, ARPACK etc. written in FORTRAN. Using these libraries in modern programming languages such as C ++ or Object Pascal is hard due to significant problems associated with the methods of passing parameters to subroutines, as well as to the method of representation of an array in memory. Another serious problem is the numbering of the elements that start from 1 in FORTRAN. In Object Pascal considering this condition may require careful analysis and the modification of the algorithm for solving the problem. This greatly increases the complexity of programming and leads to hidden, difficult to detect algorithmic errors. To provide transparent access the routines of FORTRAN-libraries while programming in Object Pascal in Delphi environment the authors developed two classes of objects for representing one-dimensional arrays TVector and two-dimensional arrays TMatrix.

DOI: 10.7256/2306-4196.2013.1.8308

Abstract: The article describes the main steps of the methodology component design of neural networks to process knowledge bases represented by semantic networks. The technique is based on the use of a unified neural network model and component-based approach to work with neural networks. An important element of the component design of neural networks is a library of neural network compatible components . One of the possible solutions to these problems may be the development of a technique of designing and using neural networks based on the unified model of neural networks and the component approach. Component Design technique is based on the use of the library of the neural networks compatible components. It is concluded that the use of the proposed methodology of component design approach will ease the design and development of the neural networks, lower qualification requirements for the developer (the end user), as well as solve the problem of neural network integration with other methods of representation and processing of information in the development of intelligent systems.

DOI: 10.7256/2306-4196.2013.1.8305

Abstract: This paper describes a method of making technical decisions using the theory of multi-agent systems. The structure of decision-making system, and describes the possible variations of its constituent components. Multi-agent systems are the result of the intersection theory of systems with distributed artificial intelligence systems. In co-operative multi-agent systems, the decision is made as a result of joint work, and competing - individual agents' actions. At the corporate structures of multi-agent systems have the property of self-organization. The problem of making decision is the task of selecting the best option out of many under conditions of uncertainty. The functions of the agent's behavior is divided into three parts: the first lays designer agent, the second is calculated as a result of the agent on development activities, and the third - formed as a result of learning agent with experience. Job agent in a partially observable, stochastic, sequential, dynamic, continuous and multi-agent environment is considered to be the most difficult.

DOI: 10.7256/2306-4196.2013.1.8311

Abstract: The article presents an analytical review of ways to implement encoders input residues in the residue number system and justified their selection of optimal structures. Authors evaluates instrumental and time costs considered circuit solutions. The purpose of the study is to consider all possible options encoding input signals in residual classes with the modern element base and full advantage of the system of residual classes, evaluation of hardware and time complexity of these options and the selection and justification of the best solution in terms of the criteria indicated above. Hardware cost of encoders are expressed as logical number of two-input logical elements, and to an EPROM contained in the form of its information bits in capacity. Instrumental cost of encoders on logic circuits depends on how many parts the input sequence is divided. It is concluded that it is possible to further simplify logical encoders and, as a consequence, reduce hardware expenses.

DOI: 10.7256/2306-4196.2013.1.8299

Abstract: In this paper we consider the problem of existing methods-ing, tools and technologies for construction machinery processing knowledge, dis-regarded approach to their construction, designed to solve the put-ing problem by integrating different methods and ways of solution of the problem, a common formal basis. Machine processing of knowledge of each particular system depends largely on the purpose of the system, the set of tasks. The main problem addressed in this paper is the lack of means to relatively inexperienced developer in terms satisfactory to design the machine processing of knowledge for application of intelligent systems for various purposes. Machine Design Technology knowledge processing involves the use of multi-agent architecture. In this paper, we consider two basic ways to classify agents: based on their functions and internal structure. Addresses a number of tools that provide additional opportunities for the design of machines processing of knowledge-based libraries. The article deals with the most important tools that enable the design of multi-agent systems. The paper discusses two basic ways of classifying agents: based on their functions and internal structure. Author consider a number of features that provide additional capabilities when designing knowledge processing machines on the basis of libraries. The article deals with the most important means to implement the design of multi-agent systems.