We consider a problem of parameters estimation for gaussian mixture models widely used in data analysis and unsupervised machine learning. A new model identification method based on Bayesian aproach and the principle of maximum posterior distribution is proposed. In the article we describe the method of multiextremum density function maximum definition using sampling by Metropolis-Hastings algorithm. The proposed method is compared with the traditional expectation maximization algorithm by computational experiments both on a sample synthetic data and the real one from > dataset.
Basics of the developed formal model (an OA-automata network) are presented in article. The model is intended for the description and the analysis of computational processes in the systems and applications working in dataflow programming paradigm. In the foundation of the model are the object-attribute approach to the organization of dataflow computational process, the theory of finite state machines (automata theory) and the theory of Kahn process networks (KPN). The developed model allows to describe concurrent computations in the distributed computing system in all its details: simultaneous work of several computing devices; direct data exchange between computing devices and an exchange through the common or distributed random access memory; data transmission on communication links connecting computing devices; synchronization of parallel computations, etc. The developed framework will be used in simulation modeling for an assessment of the computer system parameters and detection of conditions for deadlock in the parallel computation.
The article considers the basic stages in development of a local computer network and approaches to the choice of the base project variant basing on the methods of the decision-making theory. The authors propose the solution of this problem on the basis of the method of analytical hierarchy and method ELECTRE. The article describes two usage examples of these methods for the solution of a problem of types of communication lines selecting on user’s criteria.
The article considers the information structure choice of the corporate computer network based on the application of methods of decision support. The authors suggest a technique for calculation of the information structure parameters and evaluation criteria when making design decisions.
This article addresses the control task for nonlinear system that can be presented as nonlinear differential system with linear structure and parameters depending on object state. Usage of quadratic cost function allows developing control with coefficients that include solution of Riccati equation in real time. The rather common way is to solve Riccati equation at rate of object functioning which requires high performance controller what is not appropriative in some applications. The method, represented in this article, is useful when system state space is compact and performance or weight of control hardware is critical. It is offered to calculate regulator coefficients in advance and to keep them in memory of the control device. Calculation of coefficients quantity and memory size depending on accuracy of calculations is provided. The mathematical simulation of aircraft controlled by such regulator was made for verification.
This article presents three different approaches to the representation of static models in the implementation of database applications. In order to compare solutions based on selected criteria of optimality was created a unified model for testing of tools development of object-oriented applications, presented in the form of a class diagram language UML. As one of the implementations of the classical object-relational mapping patterns which proposers of static models in the environment of a relational DBMS. An alternative solution, a widely used one of the authors, is the object-attributive approach. The paper presents the main components of this approach and example of the unified model testing. The paper also presents a solution based on a matrix universal object-relational database, the main components of which implementation issues are presented in the following sections.This paper is the result of years of research and is based on numerous reports published in theProceedings of the International Scientific and Practical Conference "Object Systems" (objectsystems.ru).
Simulating principles of proposed attribute (A) and object-attribute (OA) architectures of computer systems (CS) that implement the dataflow execution model, and the results of a dataflow-supercomputer system simulation are described. A new formalism of "Attribute Nets" (A-nets) is proposed for mathematical modeling of dataflow-CS that differs significantly from apparatus of Petri Nets. This formalism lays foundation for the OA-programming&simulation environment of a dataflow-CS which is used in development programming and test of the OA-supercomputer system.
Problems of data compression are under consideration. Different modifications of the Byte Pair Encoding (BPE) algorithm are discussed. A comparative analysis of the obtained results is made from the point of view of the set of factors — the compression ratio of data and the time spent on the coding.
The paper considers the problem of reducing the dimension of the feature space for describing objects in data analysis problems using the example of binary classification. The article provides a detailed overview of existing approaches to solving this problem and proposes several modifications. In which the dimensionality reduction is considered as the problem of extracting the most relevant information from the characteristic description of objects and is solved in terms of the Shanon's entropy. To identify the most significant features information criteria such as crossentropy, mutual information and Kullback- Leibler divergence are used.
Existing program systems of project management include rather big set of the basic functionality connected with project implementation. However they don't consider specifics in work implementation, connected with research and developmental realization. As а result, based on existing programs it is necessary to develop a number of the subsystems built. These subsystems will allow to resolve problems connected with purchasing, tracking and usage after project works completion of specific equipment and materials. Besides technical questions connected with subsystems creation for work with the equipment, universal socially psychological phenomenon of enterprise labor staff incorporation is considered.