Count-Based Exploration via Embedded State Space for Deep Reinforcement Learning

Figure 4

Mean average return of different algorithms on Rllab tasks with sparse rewards.