Re:The neural network of the Stockfish chess engine
by blt on Wednesday January 13, 2021 @05:48PM
(#25769069)
From the article:
> it’s a simple feedforward network with... a single scalar output to give a numerical score for the position, indicating how favourable it is for the player about to move.
The search algorithm proposes moves, computes the board state resulting from them, and evaluates each state.
This may seem like a trivial distinction, but it's important. The Stockfish method requires knowing the rules of the game; networks that output a move directly do not.
> it’s a simple feedforward network with... a single scalar output to give a numerical score for the position, indicating how favourable it is for the player about to move.
The search algorithm proposes moves, computes the board state resulting from them, and evaluates each state.
This may seem like a trivial distinction, but it's important. The Stockfish method requires knowing the rules of the game; networks that output a move directly do not.