Difference between revisions of "Neurone"

From Chessprogramming wiki
Jump to: navigation, search
Line 24: Line 24:
=Forum Posts=  
=Forum Posts=  
* [http://www.talkchess.com/forum/viewtopic.php?t=33074 Engine update: Neurone I] by [[Alex Brunetti]], [[CCC]], March 04, 2010
* [http://www.talkchess.com/forum/viewtopic.php?t=33074 Engine update: Neurone I] by [[Alex Brunetti]], [[CCC]], March 04, 2010
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=50942&start=47 Re: Which are the chess engines written from scratch ?] by [[Luca Lissandrello]], [[CCC]], January 17, 2014
* [http://www.talkchess.com/forum/viewtopic.php?t=61848 Neurone XXV released] by [[Günther Simon]], [[CCC]], October 26, 2016
* [http://www.talkchess.com/forum/viewtopic.php?t=61848 Neurone XXV released] by [[Günther Simon]], [[CCC]], October 26, 2016

Latest revision as of 11:52, 5 April 2020

Home * Engines * Neurone

Neurone [1]

a Chess Engine Communication Protocol aka WinBoard compatible chess engine by Luca Lissandrello, written in Visual Basic .NET [2]. Neurone features standard search and evaluation techniques, such as iterative deepening alpha-beta with transposition table, and has its own endgame tablebase format with up to one pawn. It supports the Beowulf opening book format by Colin Frayn, and is able to ponder. As one of the weaker chess engines around [3], Neurone played the IOCSC 2010, IOCSC 2011, and IOCSC 2012 editions of the Italian Open Chess Software Cups, and the IGT 2013, IGT 2014, IGT 2015, and IGT 2016 over the board, and on-line the CCT14 and CCT15.


Since version XXI the ability of learning was implemented, due to the similarity with human reasoning dubbed Neurolearning. It combines supervised learning and reinforcement learning techniques, separated for the three game phases, opening (Phase 1), middlegame (Phase 2), and endgame (Phase 3) [4].

  1. Phase 1 learning considers a WDL discretized score (> 1.2,< -0.6) of the position ten moves after playing the last book move along with the final outcome as reinforcement signal, encoded in an opponent specific book learning file supplementing the standard read-only opening book.
  2. Phase 2 learning utilizes a persistent hash table, whose entries might be modified in an analysing supervisor mode, looking for possible better moves.
  3. Phase 3 learning is intended to enlarge Neurone's specific endgame tablebases.

Forum Posts

External Links

Chess Engine



  1. Neuron in a scanning electron microscope, Image by Nicolas P. Rougier, Wikimedia Commons
  2. Neurone - G 6 (Italian)
  3. Luca Lissandrello: Penso sia uno dei chess engines più deboli in assoluto… per ora. Lissandrello Luca « G 6
  4. Neurolearning « G 6 (Italian)

Up one Level