With this app, you can play and study with a CPU trained by deep reinforcement learning, which is a state-of-the-art method. The parameters are trained by the same learning method as AlphaZero shown by DeepMind, and it boasts performance comparable to top software if sufficient search is included.
In addition, the model used in this app has an added function to display the evaluation values as a distribution, so you can grasp the stability of the evaluation values at a glance. Whether the evaluation value is confidently given, or whether it is a value among many possibilities, such information may be useful in human interpretation.
C ++ programs for personal computers are available on GitHub.
https://github.com/SakodaShintaro/Miacis
In addition, the model used in this app has an added function to display the evaluation values as a distribution, so you can grasp the stability of the evaluation values at a glance. Whether the evaluation value is confidently given, or whether it is a value among many possibilities, such information may be useful in human interpretation.
C ++ programs for personal computers are available on GitHub.
https://github.com/SakodaShintaro/Miacis
Show More
