credit Mitchell Riley: we can drop 65 parameters for about the same strength; also train for longer

The idea is that the current model is M_theta(board, ply) = "score for player 0", but we can instead create a different model M'_theta(board) = "score for player 0 assuming that it's player 0's move" and for a certain pair of functions sigma, normalise, recover M_theta(board, ply) = sigma(ply) M'_theta(normalise(board, ply)).

This pair of functions is sigma(ply) = +/- 1 (parity), and normalise being a conditional interchange of pieces on the board depending on the parity of the ply. The only things stopping this from being completely straightforward is that nn_train.py evaluates M'_theta itself so we need to pass sigma(ply) to the training loop.

6x6 is still untrained.

1 files changed, 0 insertions, 2 deletions

diff --git a/include/tak.h b/include/tak.h
index 324c41f..43dd1bd 100644
--- a/include/tak.h
+++ b/include/tak.h
@@ -55,8 +55,6 @@ typedef struct tak_state_s {
   enum WIN_TYPE won;
   enum COLOUR current_colour;
   uint8_t white_count, black_count, ply, board_size;
-  // incremental zobrist hash; maintained by action_take/action_undo and
-  // re-anchored at the search root (see zobrist.h)
   uint64_t hash;
 } *tak_state_p;