[1I5-OS-31b] OS-31

[1I5-OS-31b-03] Study for the machine learning method to evaluate the situation in the simulation space

〇Atsushi Kobayashi¹ (1. Daiichi Institute of Technology)

Keywords:simulation, multi-agent, deep learning

In recent years, with respect to various digital games, there is a method of artificial intelligence that is similar to human ability and more. this method is deep learning, which performs highly in image recognition, time-sequence analysis, or natural language processes. Especially, in the game field, the deep reinforcement learning method is often applied and a gameplay result is more than human performance. However, there is a possibility that humans will act in ways that humans deem inappropriate or have difficulty interpreting the thinking process. On the other hand, a Language model is applied to the digital game field and gets high-performance results. we suppose that this model makes it possible to interpret the thinking process. In this study, the construction of a machine learning model to apply to the agent model in the simulation space is the goal. In this report, the preliminary evaluation goal is to propose the agent model to apply the machine learning method with a time sequence analysis process like the language model in the simulation space.