OutIntSys - A Novel Method for the Detection of the Most Intelligent Cooperative Multiagent Systems

Abstract

The increased intelligence of a computing system could allow more efficient and/or flexible and/or accurate solving of problems with different difficulties like: NP-hard problems, problems that have missing or erroneous data etc. We consider that even if there is no unanimous definition of the systems’ intelligence, the machine intelligence could be measured. In our research, we will understand by intelligent systems the intelligent cooperative multiagent systems (CMASs). Even in a CMAS composed of simple agents an increased intelligence emerges many times at the system’s level. We propose a novel method called OutIntSys for the detection of the systems which has a statistically extremely low and extremely high intelligence, called systems with outlier intelligence, from a set of intelligent systems that solves the same type(s) of problems. The proposed method has practical applicability in choosing of the most intelligent CMASs from a set of CMASs in solving difficult problems. To prove the effectiveness of the OutIntSys method we realized a study that included six intelligent CMASs with similar type of operation, composed of simple computing agents specialized in solving a difficult NP-hard problem. OutIntSys does not detect any outlier intelligence. It detected just CMASs whose MIQ is further from the rest but that cannot be considered as outliers. This was expectable based on the fact that the CMASs operation was very similar. We performed a comparison with two recent metrics for measuring the machine intelligence presented in the scientific literature.