Hacker News new | past | comments | ask | show | jobs | submit login

Thanks!

It would really be interesting to see how AlphaZero style game play would work out. In particular, the very wide branching factor (when accounting for stochasticity) and relatively shallow game tree might change the characteristics noticeably.

One thing that might be hard is to construct a reasonable game state representation for the neural network that is efficient for learning.




That's why it would be a fun comparison!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: