Orthogonal Evolution of Teams: A Class of Algorithms for Evolving Teams with Inversely Correlated Errors

Abstract

Summary. Several general evolutionary approaches have proven quite successful at evolving teams (or ensembles) consisting of cooperating team members. However, in this paper we demonstrate that the existing approaches have subtle, but significant, weaknesses. We then present a novel class of evolutionary algorithms (orthogonal evolution of teams (OET)) for evolving teams that overcomes these weaknesses. Specifically it is shown that a typical algorithm from the OET class of algorithms successfully generates team members that have fitnesses comparable to those evolved independently and that have inversely correlated errors, which maximizes the teams' overall performance. Finally it is shown that the OET approach performs significantly better than the standard evolutionary approaches.

Environmental robustness in multi-agent teams

“…Previous research has confirmed that OET produces teams whose average members are better than the average member in a team algorithm, teams whose members cooperate better than in an island algorithm, and teams that perform better than teams generated via either the island or team algorithms [23]. Further preliminary results suggest that OET has several significant advantages when applied to multiagent systems.…”

“…each separate island) will produce agents well suited to a particular role, and that when combined they will naturally cooperate to cover the entire problem domain. However, research has shown that the members generated tend to have significantly "overlapping" behaviors such that much of the problem domain remains unaddressed and overall performance of the team is sub-par [11,9,23].…”

Evolutionary optimization of cooperative heterogeneous teams

“…These orthogonal views of the population lead to the name Orthogonal Evolution of Teams (OET). 15 In previous work we showed that OET produces teams whose members perform better than those generated with team approaches and who cooperate better than those generated using island approaches. 15,16 However, this research was limited to classification problems where an expected failure rate model could be directly applied and where agents cooperated via a voting mechanism on classification problems.…”

Evolution of team composition in multi-agent systems

“…Thus, having the ability to discover the optimal composition is an important feature of any successful algorithm for creating cooperative, heterogeneous teams. This research extends the previously successful heterogeneous team evolution algorithm, Orthogonal Evolution of Teams (OET) [15] by optimizing team compositions as well as team cooperation.…”

“…This approach assumes that each evolutionary process will produce agents well suited to a particular role, and that when combined they will naturally cooperate to cover the entire problem domain. However, research has shown that the members generated tend to have significantly 'overlapping' behaviors such that much of the problem domain remains unaddressed and overall performance of the team is poor [7,5,15].…”