Forward-Backward Reinforcement Learning

Ashley D. Edwards, Laura Downs, James C. Davidson

Goals for reinforcement learning problems are typically defined through hand-specified rewards. To design such problems, developers of learning algorithms must inherently be aware…