[1511.09249] On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models

https://arxiv.org/abs/1511.09249