lightvector
Future Katago training - what's next?
Will there be a model succeeding b28c512 in the forseeable future? Furthermore, just out of curiosity, how are the parameters for block count and channel count determined for models?
- Duplicate of #1031
- Although varying in many minor details, KataGo’s overall architecture resembles the AlphaGoZero and AlphaZero architectures[^1][^2]
[^1]: David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, et al. A general reinforcement learning algorithm that masters chess, shogi, and go through selfplay. Science, 362(6419):1140–1144, 2018. [^2]: David Silver, Julian Schrittwieser, Karen Simonyan, Ioannis Antonoglou, Aja Huang, Arthur Guez, Thomas Hubert, Lucas Baker, Matthew Lai, Adrian Bolton, et al. Mastering the game of go without human knowledge. Nature, 550:354–359, 2017
