2024 year, volume 28, issue 1 (PDF)

Kaledina E.A., Guskova I.M. Stabilization of a controlled multiply connected system using a genetic algorithm

In this work, a multiply connected controlled dynamic system with local control is reviewed. Based on a genetic algorithm, a method for constructing local controls that stabilize the system as a whole has been developed. In this case, the parameters of the genetic algorithm are used, namely the selection, crossover, mutation operators, the type of fitness function and the critical basis. Modeling and analysis of the results obtained were carried out.

Keywords: genetic algorithms, multiply connected dynamic system, local control, asymptotic stability, stabilization.

Kozlov V.N. Electronic digital signature based on codes, defining images up to affine transformations

The first and long-standing variant of protecting a document from forgery (it is still used today) is the so-called “living” signature (or facsimile), and a clerical seal. However, nowadays the document flow is mostly electronic, and often with a very large number of documents (electronic trading, bank payment systems, transactions in crypto-currencies, etc.). The digital signature that emerged more then forty years ago works here. As a rule, the core of a digital signature is a function whose value is easily calculated for a given argument value, and the reverse, i.e. calculating the value of an argument given the value of a function is very difficult. The article describes an analogue of a digital signature on a different fundamental basis, using image codes that define them up to affine transformations.

Keywords: digital signature, image, image code, affine transformations, image authentication.

A.V. Chechkin Five principles multi-agent automated artificial intelligence systems

Goal. To investigate the effectiveness of the interaction of natural and artificial intelligence (AI) of the Earth’s noosphere. Methods. Mathematical, informational and virtual modeling. System and ultrasystem analysis and synthesis. Results. The necessity of using the concept of a multi-agent automated system as a very typical human-machine AI system of the Earth’s noosphere is substantiated. Five conceptual principles are formulated and substantiated: the principle of automation and human-centrism of AI; the principle of information system redundancy and security of AI; the principle of reasonable integrity of AI; the principle of abnormal goal setting of AI; the principle of five levels of language development of AI.

Keywords: natural and artificial intelligence, automated system, multi-agent system, digital twins, intellectual planning and situational group management system, interface language.

Kovalyova E.S. Emotion recognition algorithm based on linear regression

The paper proposes a method for recognizing emotions in facial images using linear regression using two neural networks (MediaPipe[1] and Dlib[2]), which locate key points in images of people’s faces. This article will show: creating a database for the training set, features that will be used to train the classifier, building a linear classifier and the result of its work on the training set. The goal of the work is to identify key features at key points by which one or another emotion can be recognized. The proposed approaches may be useful, in particular, when training multilayer neural networks, the training of which will improve the quality of recognition of emotions in images.

Keywords: keypoint, MediaPipe, Dlib, linear classifier.

Faershtein I.S. On the uniqueness of the solution of a system of polynomial equations over a finite field

A system of polynomial equations of a special form over the field GF(2m) is considered and the uniqueness of its solution up to a permutation of variables is investigated.

Keywords: system of polynomial equations, finite fields, uniqueness of solution.

Shergin I.A., Ryjov A.P. The problem of human pose estimation: tasks, methods, solutions

The article deals with the problem of estimating a person’s pose, which arises in many applications of computer vision systems. Various problem statements and approaches to solving the problem are considered, an overview of existing solution methods and data sets is provided. Additionally, the application of methods in the task of teaching physical skills is investigated.

Keywords: machine learning, deep learning, human posture estimation, computer vision.

Maslenikov D.O. About number of states of automaton modifying its diagram

Definitions of a rule and a result of applying it to а finite automaton are introduced. The present paper considers the result of applying the rule to the automaton as a boundedly deterministic function and a finite automaton that realizes this function. Equivalence of two definitions of a result of applying a rule to an automaton is proved. The paper provides estimations of largest and smallest number of states of the automaton.

Keywords: finite automaton, self-modifying finite state machine, Moore diagram.

Sergey Rodin On the linear realizability property of the mappings

This paper studies the property of linear realizability of mapping of the finite set into itself. This property is important from linear realizability of automata, namely linear realizability of the elements of the generating set of the automaton inner semigroup is the one of the necessary conditions for linear realizability of the automaton. Previously it was shown that every mapping of the finite set into itself is linear realizable via an encoding which code length is equal the finite set cardinality. In this paper this result will be improved and it will be shown that every mapping of the finite set into itself is linear realizable via an encoding which code length is equal the finite set cardinality minus one.

Keywords: Automata theory, semiautomata, transition systems, assignment, state encoding, complexity, boolean operator.