A short summary of literature on the description of fundamental methods and techniques, the implementation of schemes and computer programs for the construction of software for the implementation of selfdeveloping intelligent multimedia models is given. They are based on the concept of multiagency, development in systems of artificial origin. In addition, neural fuzzy software agent systems are used. Consider increasing the intellectual value of advanced multimedia models using methods based on agentordered methods. For an agent the functional model of the mixed type, in that clear and unclear rules are used in playback modification of behavior, is built. To analyze complex factors and conditions, functional dependencies over relations are used, as well as symbolic formulas for fuzzy logic over fuzzy symbols and forms. With the help of transformations over attribute values, diagrams of the rules for outputting the alphabet of the calculus are described. Their essence in functional transformations. The abilities of theoretical-categorical representation of models of similar intellectual agents and further formalization of the evolutionary formation of agent groups are discussed. In the future, it is assumed to study the potential of a theoretical and formal representation of the nature of neural fuzzy elements with various modifications of their components. It is also intended to build mappings that can convert elements of one format to elements of another format. The goal of all research is to develop new transformations.
Currently, multimedia systems are complex local multi-level batch associations [
Multimedia multitasking. Intelligent systems, as one of the subspecies of
multimedia systems, can also manage information [
The study of these issues is devoted to the work of a number of domestic and
foreign authors [
The article uses intelligent systems to solve problems of modeling and building
information telecommunication systems and describes their state using
logicalmathematical dependencies. They will be called agent-oriented systems in the future
[
In the real world, when information systems are paramount, and therefore the role of the person is significantly abolished, information intelligence agents come to the fore. To solve the resulting problems, these agents are combined into groups, which also allows them to best adapt to new environmental conditions.
Information intelligent agents are widely found in the works of domestic and
foreign authors [
However, the practical implementation of these models is still quite difficult to
implement due to the growth in geometric progression of process states. To simplify the
process of building such models, you need to consider multilevel neural models of
agents with fuzzy task conditions. Such models are also self-learning and adapt well
to the conditions of a booming heterogeneous information environment [
Over time, due to the reasons associated with an increase in modifications to information communication systems, a delay in determining the outcome of intelligent agents, difficulties appear in managing neural models. All this makes us study uncertainty in the behavior of intellectual agents, patterns in the development and selfdevelopment of intellectual agents using fuzzy neural models.
Of interest is a number of studies describing technical systems of artificial origin
from the point of view of formalizing their processes of self-improvement and
selforganization. For example, modeling formal conditions of activity, where the main
function of phenomena is the function of adaptation [
The basis of these works is an approach that describes phenomena from the point of view of their biological component. The fixture function is well represented using formalized criteria. Criteria are information technology rather than biological in nature.
In addition, a new course arose aimed at considering the artificial origin of the mind, which was called "Artificial Life". It appeared in the late 80s. XX century and its goal was to describe and model self-organizational phenomena in biotechnical systems.
The basis of this direction is work in which technical and genetic systems are
described using an NK-automatic scheme [
This direction is also very developed in the Russian Federation, the adherents of which build and study biotechnical models of behavior of various forms of life.
These models are based on an algorithm for building artificial life, concentrating on the phased development of simple organisms. These organisms are not involved in the formation of mental models. Biotechnical models themselves are most often built using computer software packages in a hardware or software technical environment.
As a result, the study of artificial systems is currently an urgent task. The study and construction of such systems using neural models with fuzzy logic is especially in demand. 3
In the structure of an artificial system, there are a huge number of different modules that can both depend on and not depend on other elements of the system. All are designed to process, store and transmit information. When interacting with other elements of the system, modules can both influence and not affect the development of the first. These modules act as intelligent agents and have all the necessary properties to solve the agent-oriented problem.
Текст When describing the work of these intelligent agents, it turns out that all
external influences that appear in the system in the form of signals, signs and signs can
be considered as certain messages and presented as functional logical dependencies.
This action is called the input language of the intelligent agent. Conversely, all actions
that the agent exerts on the external environment can also be described using
functional logical dependencies. They can also be generalized into the output language of
the intelligent agent [
The resulting functional dependencies can be converted by application packages and other software methods into different physical signals. The types of signaling can be different, both synchronous and asynchronous. Also suitable is the physical encoding of the signal.
Further, based on the obtained dependencies, the architecture of the neural network is built, which can take on both a clear and fuzzy character. Input and output signals can be measured with a certain degree of accuracy and significantly simplified by describing them with simple dependencies.
But if we consider modern information systems where it is necessary to optimize behavior under environmental conditions, then it is possible to receive messages and signals with a less clear value. By drawing an analogy, these can be variables whose values are contained in fuzzy sets. Schematically, the input and output of intelligent agent messages is shown in Fig. 1.
The intelligent agent proposed for consideration is responsible for the recognition of both correct and incorrect input information and interaction with other agents. Therefore, to create a universal model of interaction of a combined intelligent agent, we use the well-known scheme of states of an intelligent agent in a multi-agent information system.
Imagine a theoretical-multiple model of an intelligent agent in the following form LI SN , C A , K , AF - this is a tuple of relational relations, in which SN – the system number, C A – attributes, K – agents associated with the subject, AF – the function of the action [17].
Pointing and descriptive attributes will be CA {C , Cˆ} , composed of which are the exact and fuzzy attribute values, respectively. In turn, for the exact value of the attribute it will be true
Ci NCi , SCi ,VCi , and for the fuzzy one Cˆ j NCˆ j , X j , j ( x), x(t ) , where, NCi , NCˆ j - the considered properties; SCi defining set; VCi - value at some time moment t ; ( x) - a membership function with a domain of definition X j ; x(t ) - an element from a set U corresponding to a given time t .
Let us denote: {R, R} - a set of elements that reflect the nature of incoming messages; {T , T} - a set of messages outgoing from an agent. This allows translating the state model of an intelligent agent into mathematical language.
Thus, for the mathematical model of an intelligent agent, sequential operations are performed on the entered sets: {C , Cˆ} , {R, R} , {T , T} . The transition states in the behavior model are specified in the set {S} by the following classes: the reception of the elements of the set {R, R} and the impossibility of reception - {R, R} .
To find out the nature of the content of messages received by the agent and the number of clear and fuzzy attribute values, we define a set of predicates {Pr} (Pr 1, Pr 2, ..., Pr ) in the state model of an intelligent agent, which in firstorder logic constitute a set of admissible predicates.
It is possible to analyze complex conditions and relationships for incorrect input information using predicate theory and fuzzy logic, respectively, formulas F and F .
To formalize the predicate calculus, we compose the alphabet of the predicate calculus K AF : A ({C , Cˆ }, {R, R}, {T , T}, {S}, {Pr}, &, , (, ), , , @, , , 0) .
The alphabet for subject variables will have the form P ( p, q, f , hA) , p , q elements of sets {R, R} , {T , T} , respectively, that is, information at the input and output, f - logical formulas of predicates Pr , as well as logic of fuzzy statements, hC - properties of an intelligent agent hC A {hC , hCˆ } .
Let us compose the axioms of the calculus in the form: hC A (0) {hC (0), hCˆ (0)} hC (0) NC1 , SC1 ,VC1 (0) ; NC2 , SC2 ,VC2 (0) ; … NCn , SCn ,VCn (0) ; hCˆ (0) NCˆ1 , X 1 , 1 ( x1 ), x1 (0) ; NCˆ2 , X 2 , 2 ( x2 ), x2 (0) ; …, NCˆm , X m , m ( xm ), xm (0) . Here, at the initial operation time of the intelligent agent, VCi (0) - is the exact value of the i - th attribute at time t 0 , x j (0) is the fuzzy value of the j - th attribute at time t 0 .
To complete the construction of the predicate K AF calculus under consideration, we assign the inference rules in the form of a scheme of axioms: Ri ( Rl , Rk ) , Tj (Tp , Tq ) , Fj Fj (Fe , Fr (Wr )) , where Wr - the value of the degree with which the logical formula is true. Depending on the specified degree of truth, the fuzzy function will be active, and, therefore, used to apply the rule [18].
The number of inference rules that satisfy these schemes for each action model is diverse. For example, the laws that transfer elements from a state S0 to the corresponding states of the form 1 and 2 are determined by logical schemes
Ri , p @ S0 @ hC (0) @ q @ f and Ri p @ S0 @ hC (0) @ q @ f . Note that p @ Si @ hC ( Ri ) @ q, Ti @ f , Fi p @ Si @ hC ( Ri ) @ q, Ti @ f , Fi the operator was introduced to denote the state of refusal in processing incoming information Ri . In these schemes, it is likely that Ti and Fi ( Ti - messages at the output, Fi - a formula) - this is convenient as a variant of minimizing the number of outputting circuits [19].
The transition of states from type 1 to type 1, both with a truth table Fi and without Ri p @ Si @ hC @ q @ f it, is described by the schemes and schemes
. scheme , and the scheme
Logical
structures from 2 characterizes structures of the form and and indicate a return from state 2 to the same state. It should be noted that in the transition circuits, a stop is determined during the transition to a state S j where there is no case of a new output and, accordingly, a return to the original state S0 occurs.
As can be seen from the schemes for the inference axioms, they define calculation functions hC ( Si ) {hC (Si ), hCˆ ( Si )} that can be compared with the transformations of attribute values VC1 (t ) , VC2 (t ) ,…, VCn (t ) and x1 (t ) , x2 (t ) , …, xn (t ) .
Using the above, it is possible to divide intelligent agents into types, for example, such as active and passive, precise and fuzzy. Each of which can be primitive and nonprimitive, parametric and an agent with a shell [20]. 4
The developed approach to providing fuzzy logic methods in the development of multi-agent systems is to help create a set of intelligent agents capable of selfdevelopment and solve a common system problem, to organize effective interaction between agents at different levels of the hierarchy of information systems.
To organize a clear structure of interaction between agents according to certain rules, each intelligent agent is assigned a clear role based on its capabilities. And here it is possible to use both canonical number systems and number systems of a higher level. That is, the set of processed input information, objects and parameters can be ordered using the theory of structural calculus.
The developed model of a fuzzy neural structure allows agents to evolve, accumulate information and skills when interacting with the external environment of information, and increase their scope without including decentralized artificial intelligence.
In the algorithm for creating this model, the following roles of an intelligent agent are used, characterized by the level of artificial intelligence and the way of behavior: ─ reflection - the presence of a response to the constant movement of the environment and information coming from other intelligent agents; ─ focus on existing knowledge - the further behavior of agents in achieving the goal is based on the previously laid down knowledge about the environment and recognition of the situation when making decisions; ─ goal-setting and self-learning - the ability to accumulate knowledge, having a large amount of data in the form of a previously introduced base and a system of goals, behavior patterns and algorithms in unclearly specified conditions [21].
Thus, the role of the possibility of formalizing the input data in a clear or fuzzy way for building a functional model of a mixed type of the considered neural structures increases.
The indisputable advantage of agent-based models in solving technological and commercial problems allows them to be used in information systems to improve models of intelligent agents. This allows obtaining self-organizing and fault-tolerant intelligent information systems. 17. Tarassov, V.B., Metan, G.N: Merging neural network, fuzzy and agent oriented technologies: towards synergetic artificial intelligence In Proceedings of the second international conference on soft computing and computing with words in system analysis, decision and control (ICSCCW’2003, Antalya, Turkey, September 9-11, 2003), b-Quagrat Verlag, Kaufering (2003). 18. Vitenburg, E.A.: Software complex architecture of intelligent decision support in design of security systemfor enterprise information system. Vestnik of cybernetics 4(36), 46-51 (2019). 19. Gapanyuk, Yu.E., Zenger, A.S., Cvetkova, A.K., Kochkin, S.A., Cherkov, V.V.: A recommendation system building based on the approach of hybrid intelligent information systems. Dynamics of complex systems - XXI century, 14, no 2, 42-53 (2020). 20. Fisun, V.V.: System for preparing and making decisions in an intelligent information security management system. Innovation and investment 10, 103-106 (2020). 21. Silaev, Yu.V., Thor’, V.A.: On the creation of an intelligent system for monitoring and ensuring information security for use in automated systems. Informatization and communication 1, 119-121 (2017).