From 220a4b79e4b6d742a9ca7c278a1962b870500fbc Mon Sep 17 00:00:00 2001 From: tslil Date: Thu, 10 Sep 2026 14:12:41 +0100 Subject: 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. --- include/pv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'include/pv.h') diff --git a/include/pv.h b/include/pv.h index 06455fe..1754829 100644 --- a/include/pv.h +++ b/include/pv.h @@ -27,11 +27,11 @@ #define PV_MAX_CELLS (6 * 6) typedef struct pv_leaf_s { - float head[3]; + float head[2]; uint8_t cells[PV_MAX_CELLS]; - uint8_t current_colour; uint8_t terminal; float terminal_value; + float parity_sign; } pv_leaf_t; // Capture a horizon leaf (winner == 0xFF) or an exact terminal leaf. -- cgit v1.3.1