Research Article

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

Figure 2

Left submodule encodes into a feature . Right submodule predicts the action .