Learning in MDPs with highly complex state representations is currently
...
Learning from raw high dimensional data via interaction with a given
env...
The use of deep neural networks as function approximators has led to str...
Reinforcement learning policies based on deep neural networks are vulner...