Scientific Journal of KubSAU

Classic quantitative measure of the reliability of the models: F-measure by van Rijsbergen is based on counting the total number of correctly and incorrectly classified and not classified objects in the training sample. In multiclass classification systems, the facility can simultaneously apply to multiple classes. Accordingly, when the synthesis of the model description is used for formation of generalized images of many of the classes it belongs to. When using the model for classification, it is determined by the degree of similarity or divergence of the object with all classes, and a true-positive decision may be the membership of the object to several classes. The result of this classification may be that the object is not just rightly or wrongly relates or does not relate to different classes, both in the classical F-measure, but rightly or wrongly relates or does not relate to them in varying degrees. However, the classic F-measure does not count the fact that the object may in fact simultaneously belongs to multiple classes (multicrossover) and the fact that the classification result can be obtained with a different degree of similarity-differences of object classes (blurring). In the numerical example, the author states that with true-positive and true-negative decisions, the module similarities-differences of the object classes are much higher than for false-positive and false-negative decisions. It would therefore be rational to the extent that the reliability of the model to take into account not just the fact of true or false positive or negative decisions, but also to take into account the degree of confidence of the classifier in these decisions. In the intellectual system called "Eidos", which is a software toolkit for the automated system-cognitive analysis (ASC-analysis), we use initially proposed by its developers measure of the reliability of the models, which is essentially a fuzzy multiclass generalization of the classical F-measure (it is proposed to call it the L-measure). In this article, L-measure is mathematically described and its application is demonstrated on a simple numerical example

The creation of artificial intelligence systems is one of important and perspective directions of development of modern information technology. Since there are many alternatives of mathematical models of systems of artificial intelligence, there is a need to assess the quality of these models, which requires their comparison. To achieve this goal we require free access to the source data and methodology, which allows to convert these data into a form needed for processing in artificial intelligence. A good choice for these purposes is a database of test problems for systems of artificial intelligence of repository of UCI. In this work we used the database "Iris Data Set" from the bank's original task of artificial intelligence – UCI repository, which solved the problem of formalization of the subject area (development of classification and descriptive dials and graduations and the encoding of the source data, resulting training sample, essentially representing a normalized source data), synthesis and verification statistical and system-cognitive models of the subject area, identify colors with classes, which serve varieties of Iris, as well as studies of the subject area by studying its model. To solve these problems we used the automated system-cognitive analysis (ASC-analysis) and its programmatic Toolkit – intellectual system called "Eidos"

In this article we consider application of the automated systemic and cognitive analysis (ASK-analysis), its mathematical model – a systemic information theory and the program tools realizing them – the intellectual Eidos system, for input (digitization) of images from graphic files, synthesis of the generalized images of classes, their abstraction, classification of the generalized images of classes (clusters and constructs), comparison of concrete images with the generalized images (identification) of classes, comparisons of classes with each other and creations of the generalized images of genus of ground beetles on the basis of images of the types. The new approach to digitization of images of ground beetles based on use of a polar frame, the center of weight of the image and its external contour is offered. Before digitization of images, their transformations standardizing the provision of images, their sizes and an angle of rotation can be applied. Therefore, the results of digitization and the ASK-analysis of images can be invariant (are independent) concerning their situation, the sizes and turn. There is a successful experience of the solution of similar tasks in other subject domains. This article can be considered as a continuation of series of the works devoted to application of the automated systemic and cognitive analysis (ASK-analysis) and its program tools – the Eidos system

The article briefly discusses the following questions. The classic definition of virtual reality systems. Effects of virtual reality: effects of the reality, presence, depersonalization (modification of consciousness), a modification of the consciousness of the user, virtualization, interests, goals, values, and motivations ("reals and virtuals"). The criteria of reality in various forms of consciousness and their application in virtual reality. Virtual reality systems and criteria of reality, the principles of equivalence (relativity) of Galileo and Einstein and the criteria for virtual reality. The virtual device I / o. The author's definition of virtual reality systems. Dreaming, hypnagogic state, and virtual reality. Augmented reality and augmented virtuality. The modification of consciousness and the consciousness of the user in virtual reality. Consideration of future and pathological changed forms of consciousness that arise in systems with intelligent interfaces. Observance of moral norms in virtual reality and the consequences of failure. The risk of effects of virtual reality and the need for serious scientific study. The transfer of knowledge and skills from virtual reality to true. The transfer of knowledge and skills from virtual reality to true. Mechanisms of formation of models of the true and the virtual reality of man and the principles of their correct and meaningful interpretation. Principles and perspectives correct meaningful interpretation of the subjective (virtual) models of the physical and social reality formed by the human consciousness. The application of virtual reality systems. There is also a test for understanding of virtual reality

The creation of artificial intelligence systems is one of important and perspective directions of development of modern information technology. As there are many alternatives to artificial intelligence systems, there is a need to evaluate mathematical models of these systems. In this work, we consider a solution of the problem of identifying classes of levels of pay of employees on their characteristics. To achieve this goal, it requires free access to test the source data and methodology, which will help to convert the data into the form needed for work in artificial intelligence systems. A good choice is the databases from the site: http://allexcel.ru/gotovyetablitsy-excel-besplatno. In this work, we have used the database called "The database table of employees, payments calculation". The most reliable in this application was the model of the INF4 based on semantic appropriate measure of information of A. Kharkevich with integral criteria of "Amount of knowledge". The accuracy of the model is 0.960, which is much higher than the reliability of expert evaluations, which is equal to about 70%. To assess the reliability of the models in the ACS-analysis and the system called "Eidos" we have used F-criterion of van Ritbergen and fuzzy multiclass generalization proposed by Professor E. V. Lutsenko

The creation of artificial intelligence systems is one of important and perspective directions of development of modern information technology. As there are many alternatives to artificial intelligence systems, there is a need to evaluate mathematical models of these systems. In this article, we consider a solution of the problem of identifying classes of levels of pay to employees on their characteristics. To achieve this goal it requires free access to test the source data and methodology, which will help to convert the data into the form needed for work in artificial intelligence systems. A good choice is a database of test problems for systems of UCI artificial intelligence repository. In this work we have used data base on teaching effectiveness for three regular semesters and two summer semesters of 151 teaching assistant (TA) assignments at the statistics Department of the University of Wisconsin-Madison. The most reliable in this application was the model of the INF4. The accuracy of the model in accordance with Lmeasure made up 0,809, which is much higher than the reliability of expert evaluations, which is equal to about 70%. To assess the reliability of the models in the ASC-analysis and in the system of "Eidos" we use F-criterion of van Ritbergen and its fuzzy multiclass generalization proposed by Professor E. V. Lutsenko

Creation of artificial intelligence systems is one of important and perspective directions of development of modern information technology. As there are many alternatives to artificial intelligence systems, there is a need to evaluate mathematical models of these systems. In this work, we present a solution of the problem of identifying classes of salary levels of employees depending on their characteristics. To achieve this goal it requires free access to test the source data and methodology, which will help to convert the data into the form needed for work in artificial intelligence systems. A good choice is a database of test problems for systems of artificial intelligence of UCI repository. In this work we used the database called "Wine Data Set" from the Bank's original task of artificial intelligence from repository UCI. The most reliable in this application was the model of the INF4 based on semantic, according to A. Kharkevich, integral criteria of "Amount of knowledge". The accuracy of the model is 0,916, which is much higher than the reliability of expert evaluations, which is equal to about 70%. To assess the reliability of the models in the ASC-analysis and the system of "Eidos" we used the F-criterion of van Ritbergen and fuzzy multiCLASS generalization proposed by Professor E. V. Lutsenko (L-measure)

Adequate and effective assessment of the efficiency, effectiveness and quality of scientific activities of specific scientists and research teams is crucial for the information society and society based on knowledge. The solution to this problem is the subject of scientometrics and its purpose. The current stage of development scientometrics differs greatly from its previous appearance in the open as well as paid on-line access to huge amount of detailed data on a large number of indicators on individual authors and on scientific organizations and universities. In the world, there are well-known bibliographic databases: Web of Science, Scopus, Astrophysics Data System, PubMed, MathSciNet, zbMATH, Chemical Abstracts, Springer, Agris, or GeoRef. In Russia, it is primarily the Russian scientific citing index (RSCI). RSCI is a national information-analytical system, accumulating more than 9 million publications of Russian scientists, as well as information about citation of these publications from more than 6,000 Russian journals. There is a lot of data, so-called "Big data". The main primary scientometric indicator (based on which we build all the rest, such as the h-index) is the number of citations of the author's works, placed in the bibliographic database. This number of citations is determined by the software of RSCI using so-called "binding" which is a grammatical analysis and search in databases for works of the author, for relevant links from references in the works of various authors. However, the problem is, as experience shows, that authors make a very large number of simply incorrect and incomplete references in the reference lists, very far from standard. Currently, the software that RSCI uses does not automatically bind these invalid references, and this requires human intervention. But, centrally, to do this is not possible by experts of RSCI because of the huge amount of work, and distributed work for a large number of specialists in the field still requires a centralized moderation. As a result, the work for binding references to the literary sources is very slow and a huge amount of links is unbound. This leads to an underestimation of nanomatrices indicators of both individual authors and research teams that cannot be considered acceptable. The solution to this problem is offered by applying the automated system-cognitive analysis (ASC-analysis) and its programmatic Toolkit – intellectual system called "Eidos". This work provides a numerical example of the intellectual anchor of the real incorrect references to the works of the author on the basis of a small amount of real scientific data that are publicly available free on-line access to the RSCI

Classic quantitative measure of the reliability of the models: F-measure by van Rijsbergen is based on counting the total number of correctly and incorrectly classified and not classified objects in the training sample. In multiclass classification systems, the facility can simultaneously apply to multiple classes. Accordingly, when the synthesis of the model description is used for formation of generalized images of many of the classes it belongs to. When using the model for classification, it is determined by the degree of similarity or divergence of the object with all classes, and a true-positive decision may be the membership of the object to several classes. The result of this classification may be that the object is not just rightly or wrongly relates or does not relate to different classes, both in the classical F-measure, but rightly or wrongly relates or does not relate to them in varying degrees. However, the classic F-measure does not count the fact that the object may in fact simultaneously belongs to multiple classes (multicrossover) and the fact that the classification result can be obtained with a different degree of similarity-differences of object classes (blurring). In the numerical example, the author states that with true-positive and true-negative decisions, the module similarities-differences of the object classes are much higher than for false-positive and false-negative decisions. It would therefore be rational to the extent that the reliability of the model to take into account not just the fact of true or false positive or negative decisions, but also to take into account the degree of confidence of the classifier in these decisions. In classifying big data we have revealed a large number of false-positive decisions with a low level of similarity, which, however, in total, contribute to reducing the reliability of the model. To overcome this problem, we propose a L2-measure, in which instead of the sum of levels of similarity we use the average similarity by different classifications. Thus, this work offers measures of the reliability of the models, called L1-measure and the L2 measure, mitigating and overcoming the shortcomings of the F-measures; these measures are described mathematically and their application is demonstrated on a simple numerical example. In the intellectual system called "Eidos", which is a software toolkit for the automated system-cognitive analysis (ASC-analysis), we have implemented all these measures of the reliability of the models: F, L1 and L2

In the author's interpretation we consider concepts and methods of science, such as science, knowledge, model, gnosticism and agnosticism, the principle of Ashby, facts, empirical regularity, empirical law, scientific law, and others. We have formulated the main problem of the science, concluding that cognitive abilities of a human are limited and do not provide effective knowledge in a very large volume of data. The solution to this problem is to look at ways of automation of scientific research. Traditionally, we use information-measuring systems and automated systems research (ASNI) for this. However, the mathematical methods used in these systems, impose strict impracticable requirements to the source data, which dramatically reduces the effectiveness and applicability of these systems in practice. Instead of having to submit to the source data impracticable requirements (like the normality of the distribution, absolute accuracy and complete replications of all combinations of values of factors and their full independence and additivity) automated system-cognitive analysis (ASC-analysis) offers (without any pre-processing) to understand the data and thereby convert them into information and then convert this information to knowledge by its application to achieve targets (i.e. for controlling) and for solution for problems of classification, decision support and meaningful empirical research of the modeled subject area. ASC-analysis is a systematic analysis, considered as a method of scientific cognition. This is a highly automated method of scientific knowledge that has its own developed and constantly improving software tool – an intellectual system called "Eidos". The system of "Eidos" has been developed in a generic setting, independent of any domain and can be applied in all subject areas, in which people apply their natural intelligence. The "Eidos" system is a tool of cognition, which greatly increases the possibility of natural intelligence, just like microscopes and telescopes multiply the possibilities of vision (but in this case only if you have this possibility). The study proposes a new view of the models: phenomenological meaningful model, which is currently represented only by systemic cognitive models, and which is currently in the middle between empirical and theoretical knowledge. The system called "Eidos" is considered as a tool of automation of the learning process, providing meaningful synthesis of phenomenological models directly on the basis of empirical data