Giters

capn-freako / haskell-rl

Reinforcement learning in Haskell.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Set draw states to zero in initial probabilities.

capn-freako opened this issue · comments

commented

Currently, we initialize the state probabilities as follows:

  • Learner won: 1.0
  • Opponent won: 0.0
  • Other: 0.5

The problem is: some of those Other states are draw states in which the board is full.
Those should have their probabilities set to: 0.0.