Go

Building upon the existing GOjen framework, which provides an environment for playing Go games between humans and various machine players including artificial neural networks, tournaments of machine players, and training of neural players by self-play, the following topics could be addressed in a thesis:

• (Co)evolutionary ANN training
• Combination of evolutionary and reinforcement learning
• Recurrent network topologies
• Knowledge transfer of neural players to larger boards
• Analysis of the capacity of neural players to solve Go problems
• Board representations