Learning algorithm for an intelligent decision making system based on multi-agent neurocognitive architectures

doi:10.1016/j.cogsys.2020.10.015

Cognitive Systems Research

Volume 66, March 2021, Pages 82-88

https://doi.org/10.1016/j.cogsys.2020.10.015 Get rights and content

Abstract

The paper presents the formalism of an intelligent decision-making system based on multi-agent neurocognitive architectures, which has an architectural similarity to the human brain. An invariant of the organizational and functional structure of the intellectual decision-making process based on the multi-agent neurocognitive architecture is developed. An algorithm for teaching intelligent decision-making systems based on the self-organization of the invariant of multi-agent neurocognitive architectures is presented. Using this algorithm, an intelligent agent was trained and the architecture of the learning process was built on the basis of an invariant of neurocognitive architecture. Further research is related to training an intelligent agent in more complex behavior and expanding the capabilities of an intelligent decision-making system based on multi-agent neurocognitive architectures.

Introduction

The main fundamental problem of creating artificial intelligent decision-making systems is their inability to solve the unstructured tasks of the world on a par with humans. Existing systems do a good job of solving “narrow”, well-structured tasks; however, they are not capable of making effective decisions in the face of uncertainty and unstructured data. Within the framework of this problem, much attention is paid to the so-called upward approach to the development of artificial intelligence (AI) based on biological elements, since the human brain, having excellent flexibility, generalization and the ability to learn, surpasses modern intelligent systems. This approach includes neural networks that are actively developing and used in modern AI systems (Callan, 2001). However, these approaches, although they relate to biological modeling of AI, have no resemblance to biological neurons.

Thus, the goal of the study is to develop an AI system that would have an architectural resemblance to the brain and in which the decision-making process is similar to human.

The task of this study is to develop an algorithm for teaching intelligent decision-making systems based on self-organization of the invariant of multi-agent neurocognitive architectures.

The object of research is the process of learning intelligent systems.

The subject of research is the possibility of learning an intellectual system based on the invariant of multi-agent cognitive architectures.

Section snippets

Psychophysiology of decision making

The decision-making process by a person, as a rule, includes certain stages, which, in the case of solving various unstructured tasks, can be performed sequentially, simultaneously, in parallel or with a return to the previous stages (Karpov, 1999). The main stages of the decision-making process include:

1.
Situational analysis, including recognition, identification of a problem situation and its emotional assessment, goal setting, search for causal relationships to form possible solutions.
2.
The

An invariant of the organizational and functional structure of the process of intellectual decision-making based on multi-agent neurocognitive architecture

Principles outlined in Section 2 underlie the multi-agent neurocognitive architecture. On its basis, we propose building an intelligent decision-making system. In (Nagoev, 2013), the basic algorithms and methods of a multiagent neurocognitive approach to creating intelligent systems are described. In (Nagoev et al., 2020), the concept of an intelligent agent (IA) was introduced, which is an intelligent system based on a multi-agent neurocognitive architecture, and consists of software

Training of an intelligent agent based on an invariant of neurocognitive architecture

Consider an invariant-based learning algorithm as an example of an intelligent agent nutritional behavior. By nutritional behavior we understand the actions that an agent performs to obtain food, in our case, energy.

The goal setting function of an intelligent agent is to maximize its energy to increase life expectancy. Agent energy is spent both on the existence of the system and on the search for additional sources by sending messages or paying for multi-agent contracts. Also, additional

Software implementation of a self-learning intelligent agent

To test the presented algorithm, a software system that demonstrated the self-learning of an intelligent agent (Fig. 3) was developed.

The implementation of the nutritional behavior of IA shows in Fig. 3.

As in Fig. 2, the functional dependencies between the agents are drawn by lines. For clarity, the agents of various cognitive nodes are represented by their corresponding shape and color.

Conclusion

The formalism of an intelligent decision-making system based on multi-agent neurocognitive architectures, which has an architectural similarity to the human brain, is presented. An invariant of the organizational and functional structure of the intellectual decision-making process based on the multi-agent neurocognitive architecture is developed. An algorithm for teaching intelligent decision-making systems based on the self-organization of the invariant of multi-agent neurocognitive

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (10)

Z. Nagoev et al.
Multi-agent neurocognitive models of semantics of spatial localization of events
Cognitive Systems Research
(2020)
Anokhin P. K. (1975) Essays on the physiology of functional systems [Ocherki po fiziologii funktsional'nykh sistem]:...
Callan R. (2001) The Essence of Neural Networks First Edition. - M.: Williams. ISBN...
A.V. Karpov
Methodological foundations of decision psychology: Monograph [Metodologicheskiye osnovy psikhologii prinyatiya resheniya: Monografiya]
(1999)
N.K. Korsakova et al.
Clinical neuropsychology [Klinicheskaya neyropsikhologiya]
(2007)

There are more references available in the full text version of this article.

Cited by (19)

Mapping the knowledge frontiers and evolution of decision making based on agent-based modeling
2022, Knowledge-Based Systems
Citation Excerpt :
Reinforcement learning is closer to human learning, and has been a hot research topic in recent years [59–61]. Learning algorithms can also be integrated with other methods for modeling decision-making behavior, such as neurocognitive architecture, brain science, and behavioral science, in future research [62]. Another important direction for improving agent intelligence is natural language processing.
With increasing attention paid to the application of agent-based modeling in decision-making issues, a large number of related studies have been published in management science and operational research areas. This study adopts multiple methods, including bibliometric mapping, text mining, and qualitative analysis, to comprehensively review relevant research to explore knowledge frontiers and evolution in decision-making research based on agent-based modeling (DM-ABM). This review is based on 1190 relevant journal articles from the Web of Science Core Collection dataset and 37,167 collectively cited references of the articles. The top ten most-cited studies that constitute the intellectual milestones of DM-ABM were identified. Keywords and research topics develop rapidly; recent research have paid most attention to the keywords “model,” “system,” and “simulation” and topics “learning,” “contracts,” “protocols,” and “self-learning.” The top 24 references with the strongest citation bursts were displayed to show that the area was increasingly active from 2001 to 2010. Transition points were mapped to reveal the top five studies with the highest betweenness centrality, which considerably influences knowledge evolution. Then, the top three clusters are identified as the frontier areas and analyzed by text mining, including intelligent agents, model validation, and collaborative decision making. Finally, the most recent research in this field is investigated, and four future research directions are proposed: the advanced intelligence of agents, approach to reality, group decision making, innovative modeling methodologies and diversified applications.
Multi-agent neurocognitive architecture of an intelligent agent pattern recognition system
2022, Procedia Computer Science
The concept of a pattern recognition system for an intelligent system based on a multi-agent neurocognitive architecture is presented. The general multi-agent architecture of the recognition system is given. Algorithms for frame preprocessing and figure recognition similar to the organization of recognition in the primary and secondary visual cortex are presented. The basic principles of the structural and functional organization of the event recognition process using the processes of multi-agent interaction between specialized cognitive centers as part of the visual analyzer have been developed. The implementation of such an architecture will allow not only to select objects in the image, but also to create their ontology.
Application of intelligent trajectory analysis based on new spectral imaging technology in basketball match motion recognition
2024, Optical and Quantum Electronics
Intelligent Teaching Interaction System Design Based on Multi Agent
2024, Advances in Transdisciplinary Engineering
Principles of ontophylogenetic development of artificial general intelligence systems based on multi-agent neurocognitive architectures
2024, BIO Web of Conferences
Local Binary Pattern and RVFL for Covid-19 Diagnosis
2024, Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST

View all citing articles on Scopus

^☆: The work was supported by RFBR grants №18-01-00658, 19-01-00648.

View full text

Learning algorithm for an intelligent decision making system based on multi-agent neurocognitive architectures☆