NNUE is a 4 layer (1 input + 3 dense) integer only neural network. It's just over 82,000 parameters.[1] That's a very shallow, small NN - by comparison someth

NNUE is a 4 layer (1 input + 3 dense) integer only neural network. It's just over 82,000 parameters.[1] That's a very shallow, small NN - by comparison something like EfficientNet-B1[2] is 7.8M parameters, and that's considered a small network. [1] https://www.chessprogramming.org/Stockfish_NNUE#NNUE_Structure https://www.chessprogramming.org/Stockfish_NNUE#NNUE_Structu... [2] https://proceedings.mlr.press/v97/tan19a/tan19a.pdf https://proceedings.mlr.press/v97/tan19a/tan19a.pdf