Using adaptive consultation of experts to improve convergence rates in multiagent learning