The cost of learning fast with reinforcement learning for edge cache allocation Conference Paper uri icon