Hello,
I'm part of team of grad students developing an AI to play board games. Our plan is to take the DQN reinforcement learning algorithm used in AlphaGo and apply it to the game of Gomoku. We wanted to know if it would be okay for us to deploy our AI on Board Game Arena.
Instead of training the AI in a contained lab environment, we think it would be great if we could run our AI in publicly accessible environment for anyone on the Internet to play against. This way, we'll be able to test our AI against a variety of different opponents and skill levels. Also, since the game logic is already implemented on BGA, our team can focus more on the algorithm itself.
I've already played around with the site, and it looks like we can call all the necessary functions programmatically with web sockets and HTTP GET requests.
We know that there are concerns about spam and abuse with bots, so our AI will have the following restrictions
I'm part of team of grad students developing an AI to play board games. Our plan is to take the DQN reinforcement learning algorithm used in AlphaGo and apply it to the game of Gomoku. We wanted to know if it would be okay for us to deploy our AI on Board Game Arena.
Instead of training the AI in a contained lab environment, we think it would be great if we could run our AI in publicly accessible environment for anyone on the Internet to play against. This way, we'll be able to test our AI against a variety of different opponents and skill levels. Also, since the game logic is already implemented on BGA, our team can focus more on the algorithm itself.
I've already played around with the site, and it looks like we can call all the necessary functions programmatically with web sockets and HTTP GET requests.
We know that there are concerns about spam and abuse with bots, so our AI will have the following restrictions
- * Our AI will only operate from two accounts: DeepGaming, and DeepGaming2. THe AI will also announce itself as an AI before starting any matches with human players.
* Until the AI has been properly programmed and tested, the AI will only play games against itself and will not play against any human players.
* Once deployed, the AI will not join any games unless explicity invited. The AI also will not host any public games.
* The AI will automatically choose to abandon the game if proposed by the other player or if it runs into a serious error.
* The AI will only play non ELO rated matches.