A component of a Model-based reinforcement learning algorithm, which can be learned

RECURRENT ENVIRONMENT SIMULATORS

The Predictron: End-To-End Learning and Planning