Deep reinforcement learning, which has recently attracted the interest of AI researchers, combines deep neural networks (DNNs) and reinforcement learning (RL). By approximating a function in RL with a DNN, it enables an agent to learn in a complex environment represented by low-level features such as the pixels used in a 3D video game. However, learning from low-level features is sometimes problematic. For example, a small difference in input pixels results in completely different behaviors of an agent. In this study, as an example of such problems, we focus on the viewing directions of an agent in a 3D virtual environment (Minecraft) and analyze their effect on the efficiency of deep reinforcement learning.
展开▼
