Efficiently updatable neural network

Efficiently updatable neural network (NNUE, sometimes stylised as ƎUИИ), is a neural network based evaluation function that run efficiently on central processing units without a requirement for a graphics processing unit. NNUE was invented by Yu Nasu and introduced to computer shogi in 2018.[1] A NNUE merge into the code of the chess engine Stockfish has been announced.[2]

Structure

The neural network consists of four weight layers: W1 a 16-bit integer and W2, W3 and W4 an 8-bit integer. Incremental computation and single instruction multiple data (SIMD) techniques are used with appropriate intrinsic instructions, specifically in the 2018 computer shogi implementation VPADDW, VPSUBW, VPMADDUBSW, VPACKSSDW, VPACKSSWB and VPMAXSB.[1]

gollark: Quonauts 12: Switch to hexadecimal WHEN?
gollark: you utter bipolar junction transistor :bees:
gollark: Quonauts 11: metal oxide semiconductor field effect transistor.
gollark: Quonauts 11: ts better win.
gollark: Quonauts 11: please ignore the anomalous Unicode.

References
  1. Yu Nasu (April 28, 2018). "Efficiently Updatable Neural-Network-based Evaluation Function for computer Shogi" (PDF) (in Japanese and English).
  2. Joost VandeVondele (July 25, 2020). "official-stockfish / Stockfish, NNUE merge".

See also
  • NNUE on the Chess Programming Wiki.
This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.