Dyna is an architecture for reinforcement learning agents that interleav...
In model-based reinforcement learning it is typical to treat the problem...
When an agent cannot represent a perfectly accurate model of its
environ...
When faced with the problem of learning a model of a high-dimensional
en...