Back to Recipes Homepage
recipe picture
Recipe: AILearnsToPlayFoldIt
Created by Grom 102 2307
Your rating: None Average: 4.3 (30 votes)


Name: AILearnsToPlayFoldIt
ID: 102993
Created on: Mon, 01/14/2019 - 15:48
Updated on: Tue, 01/15/2019 - 17:25

version 1.01 It is a script that use genetic algorithm to find actions in game that gain score and adjust them over time. Source code can be found here:

Adjustable parameters: **Population Size** - Number of species in population. In other words - number of top species that survive to the next generation; **Mutation Size** - Number of mutated algorithms that will be added and checked each generation; **Aliens Size** - Number of random algorithms that will be added and checked each generation; **Cross Size** - Number of crossed algorithms that will be added and checked each generation; **Number Of Algorithm Steps** - Number of actions (genes) that can perform each algorithm; **Iteration Score Threshold** - Score gain threshold for algorithm to be executed again during evaluation; **Reset World Generation Each** - Number of generations that will be tested based on the same starting position before new top position will be saved as start point; **Mutate Rate** - Probability of new gene during mutation;

Also you can check/uncheck pre-defined algorithms to start with. By default two common algorithms included.

Best For


Grom's picture
User offline. Last seen 7 weeks 5 days ago. Offline
Joined: 04/30/2009
Groups: Russian team
version 1.01

- Added ability to add custom pre-defined algorithm (through "More" button)
- Fixed pre-defined algorithm selection
- Added descriptions of used actions

Joined: 10/10/2015
Groups: Team China
cool, seems faster than

cool, seems faster than Rav3n's GAB (Genetic Algorithm on Bands)

Joined: 09/24/2012
Groups: Go Science

Print actions at start before the dialog would help to know what numbers of actions to select for own algorythms.

Want to try?
Add to Cookbook!
To download recipes to your cookbook, you need to have the game client running.



Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons