Software & Systems

Articles of journal № 4 at 2017 year.

Order result by: Public date | Title | Authors

31. Synergetics of information-cognitive interaction in intelligent robotic systems with remote knowledge exchange [№4 за 2017 год]
Authors: Ulyanov, S.V., A.G. Reshetnikov
Visitors: 9185
The article describes a technology of knowledge bases remote design for fuzzy controllers, which are developed using the software toolkit “Knowledge base optimizer” based on soft and quantum computing. The paper also considers the possibility of tuning and transferring a knowledge base using remote connection to a controlled object. The presented technologies allow increasing control system robustness by adding training and adapting functions to various management situations.There is a comparison of control quality in fuzzy controllers operating in various control modes. Configuring and transferring fuzzy controller knowledge bases is performed using a remote connection with a control object online via Bluetooth and WiFi. As examples, there are different models of intelligent autonomous robots. Remote transmission of knowledge bases allows designing many different built-in intelligent regulators to implement a variety of control strategies under uncertainty and risk. The implemented technology of knowledge sharing in the group of intelligent robots allows achieving the goal of control and gaining additional knowledge by creating a new information source based on the synergistic effect of combining knowledge. The article considers various options of interaction between robotic systems. There is a brief description of each system. The experimental results demonstrate the possibility of guaranteed achievement of the control goal by a group of robots using soft computing technologies when designing knowledge bases of fuzzy controllers. The developed software toolkit allows designing and configuring complex ill-defined and poorly formalized technical systems online. This feature significantly reduces the time for intelligent control system design and improves system reliability by reducing the level of influence of expert estimates on the design process.

32. Program synthesizing based on a graph-analytic model description [№4 за 2017 год]
Authors: A.G. Zykov, I.V. Kochetkov, V.I. Polyakov, E.G. Chistikov
Visitors: 8055
The quantity and volumes of the developed software grow annually. It stimulates developers to create new tools enabling to reduce time for the next product development. It also includes testing automation equipment. The demand for new instruments of test automation increases due to increasing number of systems using different programming lan-guages. The relevance of the task of searching universal cross-language testing tools remains high. The paper considers verification of computing processes based on a graph-analytic model (GAM). The key idea of this approach is that the developed program is converted into a GAM description and is compared to the reference GAM description according to which it was created. Further, according to the results of comparing, the program either is recognized as correct, or is sent back for revision. A bottle neck of such approach is development of the program based on GAM and a potential iteration nature of the process. The authors suggest a special utility to solve this problem. This utility performs synthesis of programs for reference descriptions. The paper considers an algorithm of conversion of a GAM description object model into text representation of C# operators and expressions. A research objective is automation of program synthesis in C# by a group of GAM descriptions of a computing process. Within the research, we have created a tool enabling to transform GAM descriptions into program source codes. We have checked the developed utility on GAM descriptions of an array processing program (sorting, turn). The synthesized executed module has been successfully tested in Windows 10 operating system environment. In the future we plan to develop the utility along with new versions of a description language to enrich the possibilities of synthesizable programs.

33. Decision support system to determine a nosological entity of hepatitis [№4 за 2017 год]
Authors: Dmitriev G.A., A.N. Astafev
Visitors: 9639
The article considers the problem of determining a nosological entity of hepatitis to reduce the number of clinical errors in the assessment of medical history and clinical analysis. This problem might be solved based on iterative learning in reference situations using software that model specialists’ experience. A decision support system has a task that consists of sixty-five input factors such as tests and anamnesis and four output values that characterize the type of hepatitis. The sample of input factors is based on recommendations for a clinical diagnosis of the type of hepatitis. Nosological entities correspond to the international nosology requirements. The proposed algorithm is based on a modified iterative Kaczmarz method for solving systems of linear equations in a neural network. Solving includes calculation of weighted coefficients. This method allows performing calculations with efficient using of some computational capabilities. The system powered by Embarcadero Delphi. A database server is FirebirdSQL. The database contains 280 reference images involved in training of the system. The total amount of iterations for forming the basis of weighted coefficients is 260. The training error is 0.2–0.3. Applicability of images with the maximum error allows reducing the error and the total amount of images. The system has been clinically tested on the data of 100 real patients, the correlation index is 0.7. The developed system allows increasing the treatment efficacy in some cases. The proposed approach might be used to assess efficacy of hepatitis treatment.

34. Methods of representing text information in automated rubrication of short text documents [№4 за 2017 год]
Author: P.Yu. Kozlov
Visitors: 9040
The paper shows that citizens’ electronic messages (complaints, appeals, proposals, etc.) in terms of the possibility of their automated processing have a number of specific features. They are: usually a small document capacity, which makes it difficult to analyze it statistically, a lack of structuring, which complicates extracting information, a big number of grammatical and syntactic errors that lead to implementing several additional processing steps, thesaurus non-stationarity (composition and importance of words), which depends on the issuance of new normative documents, officials’ and politicians’ speeches, etc. All this leads to the necessity of using procedures for headings dynamic classification. The paper describes the stages of automated analysis and methods for formalizing text documents. It also proposes a developed rubrication method that uses the results of the morphological and syntactic stages with modified linguistic markup of text documents. The syntactic parser is MaltParser or LinkGrammar software that build dependency trees for all sentences in a document. The paper shows standard linguistic markings of MaltParser and LinkGrammar applied to short text documents, as well as a modification of the LinkGrammar markup to use for rubrication. Using known software for additional stages of analysis shows the problem of the diversity of linguistic markings. For example, most of the syntactic parsers at the output represent each sentence as dependency trees, which are described by linguistic markup. For further classification and assignment of weighting factors, linguistic markup should be modified, so it will increase the dimension of the metric. The developed method of rubrication takes into account the expert evaluation of the importance of words for each rubric, as well as the syntactic role of words in sentences. The paper shows a diagram of the process of automated rubrication of complaints and proposals in the developed analysis system. It also describes an experiment that confirms the expediency of using syntactic parsers in such systems, which leads to increasing accuracy of rubrication. There are recommendations to improve the accuracy of the developed method and use the theory of fuzzy sets and methods of cognitive modeling in order to solve the problem of thesaurus nonstationarity in the systems that depend on the issue of normative documents and officials’ speeches.

35. Comparison of DLA and RLA porous structure modelling algorithms [№4 за 2017 год]
Authors: A.Yu. Tyrtyshnikov, I.V. Lebedev, Ivanov S.I., Menshutina N.V.
Visitors: 10359
The paper analyzes different methods of porous materials structures generation. It shows that diffusion-limited aggregation (DLA) and reaction-limited aggregation (RLA) methods are perspective for modeling and visualization of high-porosity aerogel materials. The authors have developed algorithms for calculations using DLA and modified RLA. RLA algorithm modification in comparison with DLA is in using probability of particle aggregation when colliding with a cluster. The particle is not always aggregated, but with a given probability. This allows adjusting the branching of the generated structure. Software is implemented in C# in a Microsoft Visual Studio environment. This software requires Microsoft Windows 7 and higher operating system and at least 2 GB RAM. The paper has visualization of the generated aerogel structures. After comparing two methods it is established that the DLA method requires less computational time than RLA, however it allows modeling not only the structure but also the gelation process (structure creation), allowing controlling the specific surface area of the structure and the pore size distribution in it. The paper presents calculation data and algorithms for generating various aerogel structures. It is shown that with de-creasing the aggregation probability of the generated structure, the performance of the modified RLA algorithm decreases too. The low performance is particularly noticeable on large structures. The authors make a conclusion that it is reasonable to use the DLA method for big structure generation when it is not necessary to consider pore size distribution.

36. Fuzzy data storing and efficient processing in PostgreSQL DBMS [№4 за 2017 год]
Author: Sorokin V.E.
Visitors: 9959
In just half a century the theory of fuzzy sets has developed from theoretical foundations to practical use in artificial intelligence systems in such areas as industrial equipment and vehicles control, medical diagnosis and expert systems, including risk assessment, from economic to environmental. Many of them need efficient and reliable storage and processing of large volumes of information. For this purpose, there are DBMS, the most advanced of them are the object-relational DBMS. However, fuzzy data are difficult to reconcile with both object and relational data models. Moreover, the majority of industrial DBMS do not contain built-in fuzzy data types and mechanisms to work with them. These DBMS include cross-platform and freely distributed in source object-relational DBMS called PostgreSQL. Import substitution in infrastructure software is largely associated with application of PostgreSQL. Opportunities of fuzzy data relational modeling and a developed mechanism for extending types through the creation of the required user-defined types along with powerful procedural languages in PostgreSQL allow implementing various alternative approaches to fuzzy data storage and handling. The article shows a comparative analysis of these approaches from the point of view of maintaining data integrity and processing efficiency. There are the experimental results at various fuzzy data models depending on approaches. The paper proposes design solutions that increase the efficiency of fuzzy data search, and formulates recommendations on choosing an approach to modeling fuzzy data.

◄ ← Preview | 1 | 2 | 3 | 4