Studying Dynamics of Multi-Agent Learning in Networks

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
The British University in Dubai (BUiD)
Artificial networks mining and analysis is one of the most recent and interesting area of research due to the explosion growth of this type of networks which is seen every- where in our modern life. In the last few years few researchers’ attempts have emerged to explain the behaviors of agents in artificial networks and to optimize agent’s performance using policies learning and adaption methodology or optimizing the overall network’s performance using self-organization techniques. In most of the previous re- searchers’ attempts, studies focuses on two agents only and conducting an in-depth analysis was difficult due to lack of enhanced and improved visualization and analysis tools. In this thesis, the main purpose of the study was to investigate the dynamics of learning in scale free artificial networks (multi-agent system in particular) of different sizes, over different periods of time and using different game theory models. Using Dynamic Network Visualization and Analysis tool (DNVA); three case studies were studied. Observations post to each case study have been made and key results have been formulated based on them. The result of the study showed that different parameters used in a network for example, (size of the network, time and interaction period between agents) have significant effect on the behavior of the artificial network. Also, the study discovered the reasons behind two observations that were reported in previous research. Finally, the study of this thesis revealed issues of interest to the researchers in the filed which is the similarity in factors affecting the cooperation behavior in social and artificial networks.
artificial networks mining, network’s performance