Give a hard problem to a diverse team: exploring large action spaces

摘要: Recent work has shown that diverse teams can outperform a uniform team made of copies the best agent. However, there are fundamental questions were not asked before. When should we use or teams? How does performance change as action space get larger? Hence, present new model diversity for teams, is more general than previous models. We prove improves size gets larger. Concerning team, show converges exponentially fast to optimal one increase number agents. synthetic experiments allow us gain further insights: even though outperforms when increases, will eventually again play better large enough space. verify our predictions in system Go playing agents, where board and overcomes team.