Effective reinforcement learning following cerebellar damage requires a balance between exploration and motor noise | Synapse