In the history of infectious diseases, a single drug has been easily resisted by mutations in the pathogen. In contrast, when multiple drugs are used simultaneously, each helps prevent the emergence of resistance to the others. Ryfamate derives from Rifamate which is the combination of two medicines for tuberculosis to retard the development of drug resistance. As with this method, Ryfamate aims to combine an NNUE alpha-beta search with deep learning (DL) Monte Carlo tree search to compensate for each other's shortcomings.

Last year, Ryfamate used two NNUE functions and one ResNet function; this year, it will use one NNUE, one ResNet, and a new type of DL function, such as a deeper ResNet or Transformer. The new DL function currently has the problem of slow inference that leads to severe consequences in Shogi. However, it is considered to have a high recognition capacity. Therefore, Ryfamate will use the new function to compensate for the weak points of the previous Ryfamate with AdaBoost learning and ensemble inference.


YaneuraOu : https://github.com/yaneurao/YaneuraOu

dlshogi : https://github.com/TadaoYamaoka/DeepLearningShogi

* Ryfamate also uses a large amount of data that Kano-san, Yamaoka-san, and Tayayan-san have published.