Stateless Reinforcement Learning for Multi-Agent Systems: the Case of Spectrum Allocation in Dynamic Channel Bonding WLANs.