ISSN 0236-235X (P)
ISSN 2311-2735 (E)

Bookmark

Next issue

2
Publication date:
16 June 2019
-->

Articles of journal № 1 at 2015 year.

Order result by:
Public date | Title | Authors

1. Software for representation and transformation of discrete knowledge structures [№1 за 2015 год]
Author: Subbotin S.A.
Visitors: 7306
The article shows developed software that allows representing different knowledge models in a common language. It also contains the procedures automating the conversion between different expert knowledge. The developed program has a modular structure and contains subsystems for different knowledge representation models. The semantic network modeling subsystem contains functions for creation of the semantic network structure, adding nodes and links, deletion nodes and links, for graphic display of a semantic network with a circular, random and hierarchical arrangements of nodes, as well as a function for searching a subnetwork-request in the knowledge base network. The frame network modeling subsystem contains functions for creating frame model structure, frame creating, adding frames as nodes in the network, removing frames from the network nodes, adding and removing links between frames, graphical displaying of hierarchical frame model, and searching frames. The production model modeling subsystem contains a set of functions for knowledge creation, modifying and processing based on production models. They are: creating a model structure, adding the variable and question, adding a rule, graphical displaying of rule network, searching the basis of production model using different strategies of rule conflict resolution. The subsystem of transformation and analysis of knowledge representation structures provides a mutual conversion of knowledge from semantic networks, frame and production models.

2. Architecture of agents design support system for complex systems simulation models [№1 за 2015 год]
Authors: Pavlov A.I., Stolbov A.B.
Visitors: 6755
The paper discusses problems of automating the creation of complex systems simulation models. The authors have chosen a multi-agent simulation approach as basic modeling tool. They propose software architecture for designing multi-agent models. The main tasks of a software system are: creating conceptual model; supporting the transformation process from conceptual model into agent model; decreasing qualification requirements in programming for domain experts; collaboration of research teams.The distinction of proposed software is conceptual model explicit representation in the concept-attribute relation form and applying expert system approach to transform designed ontology into an agent model. Researchers develop the structure and behavior of agents using declarative approach. An agent declarative description is based on storing information about the agent structure and behavior separately from its implementation. The article proposes developing an agent describing mechanism allowing unified creating classes of agents that are significantly different from each other in structure and behavior. In this case a standard (unified) agent will have the following structure: states block containing agent characteristics; declarative block containing agent behavior strategy in the form of productionrules; imperative block containing software implementation of calculation procedures used in the inference process. Modern tools for expert systems design are used for declarative block implementation.

3. Tasks specification in a self-organizing informationsystem [№1 за 2015 год]
Authors: Drozhdin V.V., Shalaev A.A.
Visitors: 5612
To ensure self-modification and self-improvement, a self-organizing system should identify intra-system problems and address them via tasks formulation and accomplishment. The article presents general task specification and defines its components. As a method of solution, we suggest using semantic and constructive descriptions. The paper defines corresponding rules between a task and a method of its accomplishment and discusses the ways of forming a method that addresses the task at hand. This article highlightes the advantages of using under-defined tasks in a self-organizing system and suggests a number of ways to complete task definitions in the following cases:a) incomplete coverage of the task definition domain by the problem definition domain, b) solution method partial constraint satisfaction in relation to the initial task conditions, c) deviations from the transformations of the initial data into a solution method equivalence and from the data conversion methods that were initiated by the problem. The paper proposes the semantic task specification as a concept-task and relationships between the concepts. The concept-task consists of an identifier, a composition, content, a solving method, an external description and macro properties. Possible relationships between concept-tasks include aggregation, classification, generalization and abstraction.Semantic specifications of the tasks that are known to the system and are to be resolved in it can be used to form a layer oftasks in the “problem – task – subtask – method of solution – knowledge” semantic domain. In conclusion, the problem formulation and solution allows the system to resolve any issues that arise during the course of its operation. In addition, the use of the under-defined tasks significantly enhances system troubleshootingand makes the process more efficient.

4. Software implementation of an argumentation system withjustification degrees [№1 за 2015 год]
Authors: Vagin V.N., Morosin O.L.
Visitors: 5283
The paper gives an overview of the approaches to the argumentation formalization and considers the software implementation of the argumentation system based on defeasible reasoning. Decision support systems often contain incon-sistent and conflicting information. The methods of theclassical logics can not be applied to inconsistent knowledge bases. Argumentation is a good way to deal with such knowledge bases. Argumentation is usually considered as the process of cre-ating assumptions to solve the analyzed problem. Typicallythis process involves detecting conflicts and finding solutions. In contrast to the classical logic in the argumentation theory there may be arguments "for" and "against" certain assumptions. It is necessary to show that there are more arguments "for" than "against" some assumption to confirm it. Thus, one argument is not enough to say an assumption is plausible, but the superiority of arguments over counterarguments already does this. The paper also describes an application of justification degrees in defeasible reasoning. Justification degrees allow giving numerical assessment of argument plausibility. The authors propose the architecture of the developed argumentation system. The article presents the following methods and algorithms: the conflict search algorithm, the algorithm of new argument inference, the algorithm for calculatingjustifi-cation degrees and some others. In conclusion, the paper describes the example of the application of the system for solving a problem containing inconsistency.

5. Cognitive navigation and algorithm for creating a path text description in a convenient way for a human [№1 за 2015 год]
Author: Pestun M.V.
Visitors: 4783
A human-computer interaction when describing a route (for example, car navigator instructions) has weak expressiveness. It complicates person’s perception. Modern maps are not adjusted to a certain user knowlende anddo not allocate important information for him. It makes them less convenient. The article observes general methods of creating cognitive navigation and implementing a computer algorithm for route text description providing it in a way convenient for a human. The system considers user personal knowledge of real estate objects and organizations, user preferences, earlier routs. When a location is less familiar or unfamiliar to the user, a route description uses popular objects according to public opinion. Algorithm development is based on Google Maps system and the API library offered in a set. Google Maps system is used as a stable and proven technology which can provide good cartographical information with exhaustive information database (it is not always sufficient for this work, so theauthor used own expansion data set). However, the algorithm is multipurpose and can work on the basis of any cartographical system. This work partially relies on the researches of the person cognitive function during navigation done by the author together with the Faculty of Psychology of Lomonosov Moscow State University in CAVE virtual reality system.

6. An approach to SADT (IDEF0) difficulty evaluation [№1 за 2015 год]
Authors: Uskov A.A., Zhukova A.G.
Visitors: 5721
SADT (Structured Analysis and Design Technique) methodology and its component IDEF0 is widely used for modeling business processes, software systems, as well as technology and production processes. There is a wide range of CASE-tools to implement SADT methodology. SADT-model construction starts with the presentation of a system as a single unit and arcs representing system interfaces with the environment. It is a so called contextual diagram. Then, this block is decomposed into several blocksjoined by arcs interface. These blocks are the original function subfunctions. Each of the sub-functions if necessary can be decomposed further in a similar manner to achieve the desired granularity. The paper offers a new approach to estimating the complexity of SADT models perception based on functioning features of short-term human memory and transferring data process from short-term memory into long-term. The article considers a model perception factor and a formula to calculate it which allows comparing and optimizing SADT models for a given parameter. The paper contains a perception factor analysis for a homogeneous SADT model (a model with the same number of blocks in all the diagrams). It confirms the known empirical principle: the number of diagram blocks should be in the range 3-6. An analysis of heterogeneous 2-level SADT model has shown that to reduce a perception factor, it is necessary to choose a decomposition strategy so that the number of blocks on the charts decreased when a decomposition level increases. The authors propose to use adaptable fuzzy systems for increasing the accuracy of estimating SADT models perception difficulty. Thus, a theoretical estimate of perception factor should be used as a priori information and experimental data as a training set.

7. Software system of distributed design of complex VHDLobjects [№1 за 2015 год]
Authors: Afanasev A.N., Афанасьев А.Н., Khorodov V.S.
Visitors: 4277
The authors propose the architecture and describe the technologies of a distributed design system of complex VHDL objects. They also present the main functional modulesof subsystems and their interrelationships. A multi-agent approach is a basis for system organization, eight types of role agents are developed. The proposed description of existing agents which allows us to understand the role and place of each agent in the work of the whole system as well as the formation of a full design solution. A methodological basis of the design includes structural-functional linguistic models which are formed from the designed device VHDL description. The paper describes the program structure peculiarities on VHDL language to create a library of objects, solve design tasks by several designers in parallel, use objects from other projects and test projects with the same methodology. The article represents a design process in the system from development of designed devices specifications to saving design decisions in the knowledge base. It also describes the concept of MVC template in the form of sequence diagrams to represent functional authorization and broadcasting code on VHDL hardware description language in the structural and functionallinguistic model. The proposed system provides the development and filling of VHDL programs library which allows reusing design decisions with data modification according to requirements for new tasks. Implementation of this system allows making design more efficient due to the transfer of routine operations (search and synthesis of the design solutions) to agents as well as improving the quality of process control of collective design and reducing development costs.

8. Using possibilities of Intel MKL math library in parallel programs in T++ language for a T-system with an open architecture to improve their performance [№1 за 2015 год]
Authors: Roganov V.A., Kuznetsov A.A., Matveev G.A., Osipov V.I.
Visitors: 6107
The problem of parallel computing is a problem of software. The most common approach to developing parallel programs is based on using software packages such as MPI(Message Passing Interface). This approach requires a large amount of knowledge from developer. Developing and debuging parallel programs is also time consuming. The PSI RAS has developed a T-system which implements automatic dynamic parallelization at runtime. Input language for T-system is T++, and applications developed for the T-system are T-applications or T-programs. This paper provides an overview of products and components of Intel Math Kernel Library (Intel MKL), which contains a large set of mathematical functions and can be used in parallel T-applications to accelerate computation and get maximum performance. The paper considers the modes of using the Intel MKL mathematical library on clusters with Intel Xeon processors and one or more Intel Xeon Phi accelerators (coprocessors) from Intel company. There are demo examples of using the library in C and T++. The paper shows how the use of Intel MKL math library affects the efficiency of parallel T-programs. All experiments has been performed on “RSC Tornado SUSU” energy-efficient supercomputer of the South Ural State University.

9. Design principles for a converter that converts MS DOS projects into MS Windows projects [№1 за 2015 год]
Author: Zenkov V.V.
Visitors: 5747
Created in the past to work in the MS DOS operating system and works to this day program products gives us-ers and developers desire to reprogram them using modern tools MS Windows or another operating system. We are talking about major projects to be transferred to the new operating environment and new tools mental tools: data-base and development tools of geographic information systems.The main thing to consider when reworking the project - whether there is a new operating system tool, which is made using the old design. Working in MS DOS common high level programming languages have converters for converting source programs on modern programming languages C++ and Visual Basic. With their help, alteration projects is greatly simplified. However, the old project can be executed in a certain tool environment that does not have a converter that converts the project for the new operating system. Then reprogramming alternative project to one of the modern languages will be converter designs cre-ated by the tool used in their design. This is not about creating a converter for stacerned tool that outlived his own, and to de-velop a converter projects created with the aid of means of. Converter converts old projects and project funds of the old tool used in the old projects. When you create a converter projects requires knowledge of tools used to create the old project, language (s) program-ming language, which focuses on the converter, and those modern databases and geographic information systems to be used in new projects. This work is interesting, time-consuming andrequires a high enough skill performers. Its implementation is useful when a fairly large amount of convertible projects. This article is dedicated to the principles of the development projects of the converter based on the experienceof devel-opment of the converter designs created with the help of the fragments. The task of the converter is to automatically create a project file itself, move the old database to the new (anew database) and old data files to the new files, the old dialog forms converted into new forms. In the beginning of the need to choose a new language and a new database. If the old project hasthe functions of GIS, in addition to choose from the existing GIS.

10. Parallel algorithms for integer lattices basis reduction [№1 за 2015 год]
Authors: Kuzmin O.V., Usatyuk V.S.
Visitors: 7171
The article presents a multithread heterogeneous (GPU+CPU) implementation of block Korkin–Zolotarev lattice reduction methods. The paper considers connectionand hierarchy of the main three types of lattice basis reduction: size reduced (weak Hermit), Lovasz condition (Lenstra–Lenstr–Lovasz), Korkin–Zolotarev as known as Hermit–Korkin– Zolotarev (BKZ-methods). The authors provide references to applications in: information theory (decoding of coding group in MIMO), calculus (minimize of the positive quadratic form),complexity theory and cryptanalysis of Merkle–Hellman cryptography (solving subset sum problems), algebra and control theory (linear diophantine equation solving system), compiler theory (lattice based memory allocation), Vinogradov–Korobov number-theoretic method in the calculus, methods synthesize cryptographic and cryptanalysis in lattice basedcryptography. The authors have estimated experimental bounds of the necessary number precision to solve the shortest basis problem as well as bounds of the shortest basis problem and shortest vector problem solution accuracy under block Korkin–Zolotarev reduction. They decompose Kannan algorithms with extreme pruning techniques for solving the shortest vector problem and QR-decomposition of lattice basis. Implementation is based on POSIX Tread Library, CLAMDBLAS, CUBLAS, CLMAGMA, ACML, IMKL and shows linear speedup with a number of processor admitting variable-precision arithmetic. There is a comparison with several multithread implementations: fpLLL, CUDAEnum and parENUM. Thealgorithm implementation won the first place in the ideal lattice challenge.

| 1 | 2 | 3 | Next →