General Notes

Making a neural network setup that can play 2048 well (or at least as well as a semi-skilled human).

Goals

• (^ above)

Tweaking Parameters:

• Generations
• Nets
• Games
• Hidden Size
• Survivor Choice Method
• Mutation Rate (random boolean to determine whether or not to mutate)
• Mutation Method (different distributions, different std dev's)
• Random insertion of newcomers (yes/no, possibly how many/how often)
• Cross-mutation (two parents instead of the current one)
• Depth of game tree