Software & Systems

Journal articles №2 2018

1. Technology of interaction of IACPaaS cloud platform services with external software [№2 за 2018 год]
Authors: Gribova, V.V. (gribova@iacp.dvo.ru) - Institute of Automation and Control Processes Far Eastern Branch of RAS, Ph.D; F.M. Moskalenko (philipmm@iacp.dvo.ru) - Institute of Automation and Control Processes Far Eastern Branch of RAS (Senior Researcher), Ph.D; V.A. Timchenko (vadim@iacp.dvo.ru) - Institute of Automation and Control Processes Far Eastern Branch of RAS (Senior Researcher), Ph.D; Fedorischev L.A. (fleo1987@mail.ru)) - Institute of Automation and Control Processes Far Eastern Branch of RAS, Ph.D;
Abstract: Modern cloud technologies have a number of advantages over other types of software (simplification of installation, maintenance, access, teamwork, etc.). At the same time, some tasks are impossible to perform on cloud platforms (or non-optimal to perform). Therefore, there is an obvious need in creating standards, interfaces, systems that allow combining capabilities of different software types. This article discusses the architecture and technology of interaction between the cloud platform IACPaaS and external traditional software. The article describes the interaction architecture of the cloud platform and external software, which consists of five main components. Two of the components are on the platform: a service and a proxy agent. The others are on the external system: a web server, a proxy program and external software. The paper describes the features of each component. The authors also present a service development technology taking into account the given architecture. It consists of four main stages: 1) service development on the platform, 2) external software development, 3) web server installation, 4) communication software development – a proxy program. The features of each stage are described. The paper considers the necessary requirements for the development of agents on the platform and the specification of parameters when sending and receiving messages. There are corresponding examples. The paper gives the examples of successful use of the proposed technology when solving transport modeling problems. In the examples, the main computing load is focused on the external software, whereas services for visualizing the results of these calculations are provided on the IACPaaS platform. One example demonstrates the phased implementation of the technology. At the end of the article, there is a discussion of the advantages gained from applying the presented technology.
Keywords: technology, service, cloud platform, agents, transport modeling
Visitors: 8790

2. On the implementation of machine learning tools in real-time intelligent systems [№2 за 2018 год]
Authors: Eremeev, A.P. (eremeev@appmat.ru) - National Research University “MPEI” (Professor), Ph.D; A.A. Kozhukhov (saaanchezzz@yandex.ru) - National Research University “MPEI”; V.V. Golenkov (golen@bsuir.by) - Belarusian State University of Informatics and Radioelectronics (BSUIR) (Professor), Ph.D; N.A. Gulyakina (guliakina@bsuir.by) - Belarusian State University of Informatics and Radioelectronics (BSUIR) (Associate Professor), Ph.D;
Abstract: The paper analyzes the methods of reinforcement learning in terms of using them in real-time intelligent systems (RT IS) on the example of the real-time intelligent decision support systems (RT IDSS). The authors describe implementation of reinforcement learning algorithms based on temporal differences and consider the main advantages of using flexible (anytime) algorithms that can have a significant impact on RT IS efficiency and productivity. Flexible algorithms can be crucial to RT IDSS, as they are able to find acceptable solutions under tight time constraints and improve them (up to optimal ones) while increasing available resources (especially temporary ones). The proposed flexible algorithm includes a statistical forecasting module and a module of multi-agent reinforcement learning. The paper considers the possibilities of implementation of the developed flexible algorithm into a RT IS forecasting subsystem of RT IDSS type to control and monitor a complex technological object. The paper considers the approach to implementation of a transition from knowledge-based intelligent systems training to training of their development tools. At the same time, the architecture of such intelligent systems is considered as the basis for its flexibility and learning capacity. The paper also examines the areas of intelligent system learning and self-learning, as well as their ability to acquire knowledge and skills from different sources. The authors provide rationalization for the application the developed OSTIS technology to develop knowledge-based intelligent systems, including the RT IS.
Keywords: artificial intelligence, development technology of intelligent systems, anytime algorithms, time, reinforcem ent learnin, self-learning, machine learning, decision support, real time, intellectual system, software
Visitors: 11015

3. An enterprise management system [№2 за 2018 год]
Authors: Yu.M. Lisetskiy (Iurii.Lisetskyi@snt.ua) - S&T Ukraine Company (Director General), Ph.D;
Abstract: The article considers an enterprise as an open economic and social system that acts in terms of situational approach. This supposes changes of enterprise management forms, methods, systems, style depending on the objective conditions of the environment, as well as accelerated convergence of management models. It is demonstrated that management is a purposeful and continuous process of management subject influence on a controlled object, which is aimed at changing a state according to a defined plan and which is an essential function of any system. There are basic principles (cybernetic laws) of building management systems that include variety, differences of general from particular, external addition, feedback, anti-entropy, as well as cybernetic and functional models of a management system. The paper considers an enterprise management system as a form of real implementation of managerial interconnections and its composition including the subsystem of management methodology, process, structure and technique.
Keywords: enterprise, economic and social system, a control system, functions, principles, laws, feedback, structure, mathematical model
Visitors: 35686

4. Application of a knowledge representation productional model to assess compliance of an applicant training level to requirements of a position in IT department [№2 за 2018 год]
Authors: Zubkova, T.M. (bars87@mail.ru) - Orenburg State University, Ph.D; Tagirova, L.F. (LG-77@mail.ru) - Orenburg State University (Associate Professor), Ph.D; V.K. Tagirov (vladtagir@mail.ru) - Volga State University of Telecommunication and Informatics (Orenburg branch) (Associate Professor), Ph.D;
Abstract: Nowadays, due to fast development and widespread introduction of information technology in various spheres of life, a large number of specialists in this field (programmers, system administrators, etc.) appeared in the labor market. Therefore, it becomes more difficult for employers to decide on a suitable employee among a large number of potential appli-cants. Besides, employees of recruitment departments have to spend much time to entering and processing applicant data for various positions, including ones in the IT department. The paper describes the solution of the recruitment problem in the IT department based on quality assessments of applicants. There is an algorithm and a software system that allows choosing the most suitable employee for a vacant position using quantitative assessments of professionally significant and personal qualities of applicants. The paper proposes the developed software system to automate the information recruitment processes. It will allow assessing formedness of the required qualities of applicants for a position and store the results in a database for further processing. An expert assessment module was used to support a decision on whether a training level of an applicant met the requirements of a particular IT-department position. The core of this module is a product model of knowledge representation. When the software system is running, applicant information is read from the database and compared with the rules. As a result, the system gives recommendations to a decision maker on a position an applicant can aspire.
Keywords: staff recruitment, it-department, professionally significant and personal qualities of applicants, software system, expert estimation, knowledge base, knowledge representation production model, decision support
Visitors: 11312

5. Search of minimal acyclic path in weighted graphs containing negative weight edges [№2 за 2018 год]
Authors: Gdansky N.I. (al-kp@mail.ru) - Moscow Polytechnic University (Professor), Ph.D; Kulikova N.L. (kulikovanl@mpei.ru) - National Research University "Moscow Power Engineering Institute" (Associate Professor), Ph.D; E.V. Chumakova (cpp@mami.ru) - Moscow Aviation Institute (National Research University) (Associate Professor ), Ph.D;
Abstract: The paper considers the problem of determining minimal acyclic routes in weighted graphs containing negative weight edges. It also investigates the possibilities of the Bellman-Ford algorithm. It is shown that in general all its possible modifications, even the deepest ones, can determine only approximate solutions. On the average, the best approximate solutions to the problem are given by a modified version of the Bellman-Ford algorithm, in which cycles are discarded when searching for optimal continuations of the desired path. To ensure the search for exact solutions, the authors propose a modification of the branch and boundary method based on using special acyclic search trees. In the nodes of this tree, along with the number of the next vertex, the current cost of the path and the necessary pointers, there are the numbers of all the untraversed vertices, as well as a special list of negative weight edges. A list of vertices is used for acyclic continue of a path. A list of negative weight edges is necessary to estimate the maximum possible decrese of its acyclic continue. The paper suggests to use special bidirectional lists for more convenient practical implementation of acyclic search trees for one level vertices. In such lists, direct links specify the movement to neighboring level nodes. Back references point to the previous vertex in the path from the initial vertex to the given node. The algorithm is described on a C-like pseudocode. There is also an example that demonstrates the approximate character of Bellman-Ford algorithm modifications on the one hand and the work of the proposed method on the other hand.
Keywords: weighted graphs with negative weight edges, minimal route, bellman-ford algorithm, branch-and-bound method, search tree, bidirectional lists
Visitors: 9117

6. A method for determining the capabilities of parallel execution of data mining algorithm functions [№2 за 2018 год]
Authors: I.I. Kholod (iiholod@mail.ru) - St. Petersburg Electrotechnical University "LETI" (Associate Professor), Ph.D;
Abstract: The article describes the method for determining the possibility of parallel execution of data mining algorithm functions. The input parameters of the method are: a data mining algorithm represented as a composition of thread-safe functions, and a mining model represented as an array of unified element trees describing the patterns mined by the algorithm. When determining the possibilities of parallelization, the method takes into account information dependencies between functions, which are determined by a set of used and modifiable elements of a mining model. The method analyzes data connections for each pair of algorithm functions when testing the possibility of task parallelization, as well as for calling functions in loops for different iterations, while verifying the possibility of data parallelization. The analysis includes checking necessary and sufficient conditions for parallel execution for systems with shared and distributed memory. They extend the Bernstein conditions used in compiler theory and are sufficient, but not necessary. The method determines the possibilities of parallel execution of functions for shared memory systems and distributed memory systems. In this case, parallel performance of functions in shared memory systems is more efficient due to the lack of calls for the cloning function and combining of mining models. The result of the proposed method is the parallel form of the source data analysis algorithm. It contains insertions of special functions of higher order that provide parallel execution of algorithm functions that satisfy necessary and sufficient conditions. To illustrate the proposed method, the paper shows parallelizing 1R classification algorithm. It defines algorithm functions that can be paralleled, both by data and by tasks. The result is a parallel form of the 1R algorithm with parallelization function inserts for shared memory systems and distributed memory systems.
Keywords: parallel algorithms, distributed systems, data mining, data dependency
Visitors: 7004

7. A recommendation system based on data mining of a scientometric research profile [№2 за 2018 год]
Authors: Valko D.V. (valkodv@inueco.ru) - South-Ural Institute of Management and Economics, South-Ural Versatile College (Head of Laboratory), Ph.D;
Abstract: Nowadays scientific results can be represented in various scientometric bases and systems. They are often popular not because of their relevance, but due to global availability. In fact, scientific results may be out of the scope of a scientific community simply because they are not placed in a popular scientometric system. From scientific point of view, such situation devalues a researcher regardless of quality and relevance of his scientific results. According to the author, the development of recommendations for individual researchers, research teams and their managers at all levels of management would make it possible to pay attention to promising scientific results and reasonably accumulate necessary resources to include such results in popular scientometric systems. Development of tools that operate the big scientometric open data cannot be without data mining methods. The paper shows that based on the algorithm of intellectual analysis of interrelations (like apriori), which is adapted to scientometric data in the Scopus, it is possible to formulate certain sets of associative rules suitable for forecasting probable future scientific results. It is also possible to develop automated recommendations for improving publication activity. The paper proposes the developed methodical tools for analyzing publication and scientometric data using data mining methods. In addition, it describes a modular architecture and a prototype of a software system that allows forming a publication rating and individual recommendations for improving author's publication activity based on scientometric data mining. The paper shows some experimental results on assessing publication potential of researchers affiliated with the South Ural State University. The authors built necessary diagrams in the UML 2.0 notation describing that software system using Microsoft Visio modeling environment.
Keywords: scientometrics, recommendation system, data intelligent analysis, publication activity
Visitors: 8676

8. Modern approaches to training intelligent agents in the Atari environment [№2 за 2018 год]
Authors: D.A. Korobov (тroyalruby@yandex.ru) - St. Petersburg Electrotechnical University "LETI" (Student); S.A. Belyaev (beliaev@nicetu.spb.ru) - St. Petersburg Electrotechnical University "LETI" (Associate Professor), Ph.D;
Abstract: The article describes the research and development of machine learning algorithms for the problem of training intelligent agents in the Atari environment. The environment is the game console Atari2600 emulation on OpenAI Gym platform. The main aims of the research are to describe and compare modern learning algorithms with reinforcement in a number of cases, to expose their advantages and drawbacks and to give suggestions in order to increase their effectiveness. The authors describe a mathematical model of the reinforcement learning task in terms of Markov Decision Process and form a system optimality criterion. In addition, they note the Atari environment properties. There is a description of an operation principle of existing solutions, as well as the tools used in them. The paper considers a basic Q-learning approach and its well-known modification that uses a neural network. It also presents the idea of the algorithm based on policy gradient. The paper gives a description of the approach of “actor-critic” algorithms and considers its asynchronous multithreaded version. All the described algorithms are presented in pseudocode. The authors propose a method of increasing the effectiveness of learning using feature extraction. The developed algorithm for feature vector generation is based on image recognition. The paper describes the steps of feature vector selection and its usage during the learning process. The described algorithms are implemented, and the experiment is conducted. A comparative analysis of the results gave conclusions on the efficiency of the algorithms. Finally, the authors propose some ideas for further increasing the speed and efficiency of training of intelligent agents.
Keywords: atari, intellectual agents, time, reinforcem ent learnin, q-learning, policy gradient, asynchronous actor-critic, pattern recognition, feature matching
Visitors: 10104

9. A method of automated development of the semantic dialogue system database model [№2 за 2018 год]
Authors: Posevkin R.V. (rus_posevkin@mail.ru) - The National Research University of Information Technologies, Mechanics and Optics;
Abstract: The paper considers the problem of intellectual database content analysis for creating a semantic database model. Voice assistants are created to simplify interaction with mobile devices such as smartphones and tablets. A text dialogue system is an analogue of this approach. As a result, user can interact with software using natural language. User interface is a set of software solutions that helps to search, review, obtain and process information from a database that is external storage. Natural language interface is a sort of user interface that accepts and processes natural language queries. This interface can also use natural language in output to show found information to a user. A semantic database model is an important part of the dialogue system. This model includes interconnections and internal structure of a database. Manual creating of a semantic database model significantly increases time and labour costs, as well as development cost of a software system. The main purpose of the article is automation of development of a dialogue system semantic database model. The proposed method uses a set of approaches to automated creating of an existing database semantic model. An object domain thesaurus helps to define semantics and solve the problem of polysemy in text processing. Patterns helps to extract interconnections in a database. The analysis of database field content allows determining data semantic and nature. Locale indication allows decreasing the time for a database content analysis.
Keywords: natural language processing, semantic model, user interface, database, dialogue system
Visitors: 8712

10. The system for parallelizing the load on computer resources [№2 за 2018 год]
Authors: E.V. Palchevsky (teelxp@inbox.ru) - Financial University under the Government of the Russian Federation; A.R. Khalikov (khalikov.albert.r@gmail.com) - Ufa State Aviation Technical University (Associate Professor), Ph.D;
Abstract: The article is devoted to parallelization of the load on computer physical resources using an algorithm based on Markov chains. The IDLP hardware module (I distribute the load processes) is designed to parallelize computation processes to physical resources of a computing cluster. The paper shows load dependence of physical resources on the tasks being run on a computing cluster. It also substantiates the expediency of using the IDLP hardware-software module. The created hardware and software module allows efficiently handling transfers of computing processes and prevents system congestion. One of the module elements is a user-friendly interface that allows remote managing IDLP from both a smartphone and a personal computer. The load parallelization system consists of three stages. The first one is algorithm development, the second one is technical implementation, the third one is hardware-software module testing. At the first stage, the authors present the functional of the developed hardware and software module, which includes start, stop, restart, set the kernel load value. The paper presents the scheme of the developed hardware and software module. At the second stage, a source code fragment (in Python) is responsible for uniform distribution (parallelization) of computational processes on physical and logical computer cores, followed by decreasing load on a central processor. At the third stage, the authors test the hardware and software module. The testing confirmed the decrease in the workload of physical server computing resources, and, as a result, it led to increased productivity. The developed solution not only raises productivity by factor of 1.72 (on the equipment used; the load may differ in other cases), but also allows parallel running of complex and resource-intensive computing processes without disrupting computer performance.
Keywords: parallelization, process, data processing, clusters, computing process
Visitors: 9399

| 1 | 2 | 3 | Next → ►