1
Simulation / Help me... building a Genetic AI over Geronimoo
« on: June 29, 2014, 11:41:39 pm »
Hello,
I'm building a Genetic IA that will fight against a Geronimo's bot and find a strategy that beats it.
Would anyone be interested in developing it with me? Any feedback?
For now, my bot runs only against a basic BM and struggle to get more than 70% wins on a board containing only basic victories and basic treasures.
https://github.com/JonathanScher/Geronimoo-s-Dominion-Simulator
I have a running AI that needs further development:
- Optimise Geronimo and the Genetic AI
Geronimo constantly counts the number of pile depleted, and takes a long time to figure out victory points. The GAI creates a lot of new games instead of recycling games that have ended.
I would need someone to help me optimise Geronimo and my bot.
- Better GUI
I use a standard GUI that present the best bot for every step and draw a general graph.
How about choosing the opponent bot? Or giving the user the ability to save the simulation to restart it later? Or providing a way to enter the initial population size and the selection strategy?
- Code cleanup
I started with TDD, but now it's messier. How can we clean that up?
Anyone interested? Any feedback from anyone?
Cheers,
Jonathan
I'm building a Genetic IA that will fight against a Geronimo's bot and find a strategy that beats it.
Would anyone be interested in developing it with me? Any feedback?
For now, my bot runs only against a basic BM and struggle to get more than 70% wins on a board containing only basic victories and basic treasures.
https://github.com/JonathanScher/Geronimoo-s-Dominion-Simulator
I have a running AI that needs further development:
- Optimise Geronimo and the Genetic AI
Geronimo constantly counts the number of pile depleted, and takes a long time to figure out victory points. The GAI creates a lot of new games instead of recycling games that have ended.
I would need someone to help me optimise Geronimo and my bot.
- Better GUI
I use a standard GUI that present the best bot for every step and draw a general graph.
How about choosing the opponent bot? Or giving the user the ability to save the simulation to restart it later? Or providing a way to enter the initial population size and the selection strategy?
- Code cleanup
I started with TDD, but now it's messier. How can we clean that up?
Anyone interested? Any feedback from anyone?
Cheers,
Jonathan