Learning algorithm for 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)
- 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...
Methodological foundations of decision psychology: Monograph [Metodologicheskiye osnovy psikhologii prinyatiya resheniya: Monografiya]
(1999)- et al.
Clinical neuropsychology [Klinicheskaya neyropsikhologiya]
(2007)
Cited by (19)
Mapping the knowledge frontiers and evolution of decision making based on agent-based modeling
2022, Knowledge-Based SystemsCitation 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.
Multi-agent neurocognitive architecture of an intelligent agent pattern recognition system
2022, Procedia Computer ScienceApplication of intelligent trajectory analysis based on new spectral imaging technology in basketball match motion recognition
2024, Optical and Quantum ElectronicsIntelligent Teaching Interaction System Design Based on Multi Agent
2024, Advances in Transdisciplinary EngineeringLocal Binary Pattern and RVFL for Covid-19 Diagnosis
2024, Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST
- ☆
The work was supported by RFBR grants №18-01-00658, 19-01-00648.