Predecessor Features

Duncan Bailey, Marcelo G. Mattar

Any reinforcement learning system must be able to identify which past events contributed to observed outcomes, a problem known as credit assignment. A common solution to this probl…