Русский
12.2014 - volume 18 issue 4 (PDF)
This article will discuss solving the problem of recognizing illegal content in ads on the Avito website [1], published on the popular service for solving machine learning problems Kaggle [2]. We considered an algorithm based on transforming text fields using natural language processing methods [3] and finding rules from words and phrases that identify an ad as spam. The paper examined the main options for processing text information, and also proved on the basis of experiments that complex processing, such as building bigrams and stemming, does not lead to increased accuracy.
Keywords: big data, text analysis, natural language processing, data mining, association rules.
To solve a number of problems of medical diagnostics, possibilistic methods for modeling medical objects and the process of making a diagnosis, based on the modern theory of possibilities [1], have been developed and studied. The article is devoted to the application of probabilistic and possibilistic algorithms and programs for learning and recognition in the conditions of fuzzy description of medical objects and variability of their probabilistic characteristics over time. The software package for diagnosing diseases allows for the training and diagnosis procedure to be carried out in a dialog mode with the participation of a doctor and is used in this work to solve the problem of diagnosing acute appendicitis.
Keywords: pattern recognition, software package, identification problem, probabilistic diagnostic model, possibilistic diagnostic model, granulation, Kohr's classification algorithm, acute appendicitis.
This work considers the possibility of using a multi-agent approach to calculate the energy balance of the city of Kaliningrad and its neighboring countries. The stages of constructing a multi-agent model in the Anylogic simulation environment are shown, and the results of modeling are presented. The advantages and limitations of this approach and the possibilities of its application to solve other problems are discussed.
Keywords: energy, agent, model, smart grid.
For the first time, a fundamental methodology of intelligent systems of full intelligence with intuition for a wide practical purpose, capable of dynamically and effectively self-learning, self-organizing and making decisions quickly, is presented. Technical solutions for the practical implementation of an intelligent system with full intelligence and intuition are presented.
Keywords: intelligent system, full intelligence, intuition, learning, recognition of information patterns.
A system of control of the operator's condition of a complex technical system is proposed, oriented on the analysis of his speech response. To enable work in strong noise, additional channels of information input are provided, as well as a special scenario of the dialogue of the operator with the computer.
Keywords: speech recognition, strong noise, self-comparison.
A variant of the theory of possibilities is considered, which allows describing the agreement of a research group on their common interpretation of some values of possibility and necessity. This makes it possible to clarify the criterion of optimality of the decision rule in the identification problem. For the case of modeling a stochastic object, an algorithm for restoring the possibility taking values in a scale with fixed points, and these points themselves, is described.
Keywords: possibility, necessity, agreement, optimal decision rule.
This paper introduces iterative propositional calculi, which are finite sets of propositional formulas together with the modus ponens operation and the superposition operation defined by the set of Mal'tsev operations. For such calculi, the question of decidability of the expressibility problem is studied. In particular, it will be shown that there are undecidable iterative calculi. An approach to describing decidable iterative calculi will be proposed, based on defining such calculi by clones of k-valued logics. In addition, a lattice of clones of three-valued logic generating iterative calculi will be described, and the continuity of the set of decidable iterative calculi will be proved.
Keywords: iterative propositional calculus, interpretation, finite model, clones of three-valued logic.
The information monitoring technology was developed to analyze complex, weakly formalized problems (processes) based on all available information, to build forecasts of their development and to develop recommendations for managing their development. In this paper, the information monitoring technology is formalized using the classical apparatus of discrete mathematics — functional element schemes and functions of k-valued logic. This formalization solves two key problems of the information monitoring technology — checking the stability of the model and the problem of optimal resource allocation.
Keywords: information monitoring, optimal resource allocation, stability of discrete systems.
The article discusses the problems of generating specifications, characteristics and diagrams of ventilation units in dxf format, as well as generating assembled units in 3D format. This functionality should be implemented using web technologies, since it will be embedded in a web application for calculating ventilation equipment. In particular, the ability to rotate the 3D model of the unit in the browser should be implemented.
Keywords: drawing generation, AutoCAD, openjscad, 3D, dxf, ventilation units.
The article considers the task of developing an industrial ventilation equipment configurator in the form of a web application. We implement a web interface that allows you to assemble any supply, exhaust and supply and exhaust unit in the browser by simply dragging sections with a mouse or finger. This solution is ready for embedding in a web application for calculating ventilation units.
Keywords: JavaScript, HTML, CSS, configurator, industrial ventilation.
A method for solving the problem of "exponential explosion" of the number of states of a finite automaton recognizing a set of regular languages defined by the union of regular expressions of the form . \(* R_1. * R_2.*\), where \(R_1\) and \(R_2\) are arbitrary regular expressions is considered. An extension of this method is proposed for the case of a union of an arbitrary number of regular expressions of this type. Estimates are given for the number of states of the automaton with such a change in the case of an alphabet consisting of at least three symbols. It is shown that the relative decrease in the number of states can be arbitrary. The practical efficiency of the proposed method is analyzed as applied to regular expressions of the Snort system.
Keywords: finite automata, regular expressions, intrusion detection systems.
This paper studies the functioning of the lungs of a smoker in a clean environment. The dependence of the deterioration of the efficiency of cilia on the time during which nicotine was in the lungs was revealed. A time estimate of the complexity of the process of self-cleaning of the lungs from nicotine was also obtained.
Keywords: lungs, self-cleaning process, clean environment, nicotine.
In this paper, we will consider propositional calculuses whose formulas are formed by logical connectives containing classical implication, and whose inference rules are the modus ponens and substitution operations. It is known that in the general case, the problem of recognizing the expressibility of some calculuses through others is algorithmically unsolvable. In this paper, we will consider special cases of this problem: recognition of axiomatization, recognition of extension, and recognition of completeness. In particular, it will be shown that the problem of recognizing an extension is algorithmically undecidable for any calculus, and the problems of recognizing axiomatization and completeness are algorithmically undecidable for any calculus from which the formula \(x \rightarrow (y \rightarrow x)\) is derivable.
Keywords: classical and intuitionistic propositional calculus, implicative calculus, decidability, recognizing axiomatization, extension, and completeness, tag system.
The concept of stellar height was first introduced by Eggan in 1963 along with the concept of cyclic complexity of a finite automaton. The cyclic complexity of an automaton is related to the stellar height of the language it accepts: the stellar height of a regular language is equal to the minimum cyclic complexity among automata that accept this language. However, the problem of finding a minimal (with respect to cyclic complexity) automaton still remains open. This paper considers the relationship between the cyclic complexity of a minimal automaton of a regular language and its stellar height.
Keywords: finite automata, regular languages, stellar height, cyclic complexity, minimal automaton.
The problem of finding the chromatic number of graphs is one of the most attractive and challenging problems in graph theory. It is known that for biplanar graphs (graphs realized without intersections of edges on two sides of the plane), the chromatic number is not less than 9 and not more than 12. In this paper, we show that the chromatic number of biplanar graphs without triangles is not less than 5 and not more than 8.
Keywords: biplanar graph, chromatic number, triangle-free graph, graph of thickness 2.
The paper continues the study of issues of algorithmic recognition of the property of reversibility for cellular automata. A class of two-dimensional binary cellular automata with a fixed local transition function with 91 variables is constructed, in which the problem of reversibility property recognition is algorithmically undecidable.
Keywords: cellular automata, reversible cellular automata, Turing machines, halting problem.
The article presents a new proof of Gödel's theorem on the incompleteness of formal logical systems, based on concepts of functional programming.
Keywords: mathematical logic, incompleteness of formal systems, Gödel's theorem, functional programming.