During the special period of Covid-19, I joined a special online activity in Animal Crossing (a popular game on switch) which is a meeting about Artificial Intelligence discussion, and people all around the world can directly talk online, on zoom or on animal crossing. The holder will select 16 projects all over the world and they can make a presentation on animal crossing about their project.
My team mates and I decide to do an interesting project, and I wrote an abstract for our project:
Reinforcement learning called Deep Q Network is a learning algorithm that can train the NNs without Dataset. It has a prospect to becoming an efficient method to require the machines of autonomous learning ability to finish a specific task. When NNs is being trained, DQN is able to automatically adjust parameters and record the optimum data for further trainings.
In 2013, Tencent, a Chinese privately operated IT company, has introduced a mini-game called Aircraft War on WeChat, and this game has been popular for a while at that time. This dissertation considers the scenario in mini-game as a sample and examines the effectiveness of Deep Q Network while training agents to automatically play this game. In other words, the dissertation will discuss if DQN can effectively train an agent to automatically get a high mark in this game. The dissertation additionally defines distinctive observations and compares agents’ training process and performance situations under diverse observations.
The anterior researchers have frequently tested in electronic games to figure out the availability of different reinforcement learning algorithms. As senior high school students, we are trying to establish a networking platform to promote communications between the players of Aircrafts War all over the world. As soon as players enter the platform, they can experience Aircraft War under the DQN handler. Players are also able to directly change code parameters of DQN to control the agent so as to enhance their participations of this learning algorithm.
Key Words：AI Game, Reinforcement Learning, Deep Q Network
We then sent this abstract to the holder and luckily, we are selected to make presentation as one of 16.
