Question

in reinforcement learning, no labeled input is required but just like unsupervised learning there is a need for sub-optimal actions that must be explicitly corrected for learning to occur.