Page 1 of 2
neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Thu Apr 19, 2012 2:21 pm
by Jim Ablett
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Fri Apr 20, 2012 9:41 pm
by Denis P. Mendoza
Jim,
Could you add 'neurochesstrainer' to make it a complete package?
Thanks.
Denis
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Fri Apr 20, 2012 9:45 pm
by Jim Ablett
Denis P. Mendoza wrote:Jim,
Could you add 'neurochesstrainer' to make it a complete package?
Thanks.
Denis
Hi Denis,
it should all be in there.
Jim.
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Mon Apr 23, 2012 5:44 pm
by Denis P. Mendoza
Jim Ablett wrote:Denis P. Mendoza wrote:Jim,
Could you add 'neurochesstrainer' to make it a complete package?
Thanks.
Denis
Hi Denis,
it should all be in there.
Jim.
Sorry, I didn't look first before I complained!
Thanks a lot for these. I only asked as I wasn't successful with the trainer.
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Mon Apr 23, 2012 7:47 pm
by jm_thompson
Has anyone tested this Neural Network Stockfish? how does it compare with the original one? is it any stronger?
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Mon Apr 23, 2012 8:32 pm
by royb
How does this all work? I'm new to neural nets but believe that they typically have to be "trained" before playing well. What is the process for the training and how long before it begins to play reasonable (2000 ELO) chess?
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Mon Apr 23, 2012 8:39 pm
by gleperlier
jm_thompson wrote:Has anyone tested this Neural Network Stockfish? how does it compare with the original one? is it any stronger?
** neuroStock is an experiment (more like meta-engine) and in
** combination with neuroChessTrainer[fn:2] supposed to be testing
** platform for different backpropagation neural networks. There
** are no expectations for engine strength and as far as hardly
** trained, it plays like a blond bimbo.
But project is very interesting
Will try it ^^
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Mon Apr 23, 2012 9:06 pm
by gaard
I am not getting the output logs using mode 0. Could this be because I am using Windows and you compiled with the original directory format "./logs/stockfish-"?
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Mon Apr 23, 2012 10:58 pm
by jm_thompson
gleperlier wrote:jm_thompson wrote:Has anyone tested this Neural Network Stockfish? how does it compare with the original one? is it any stronger?
** neuroStock is an experiment (more like meta-engine) and in
** combination with neuroChessTrainer[fn:2] supposed to be testing
** platform for different backpropagation neural networks. There
** are no expectations for engine strength and as far as hardly
** trained, it plays like a blond bimbo.
But project is very interesting
Will try it ^^
thanks, I didn't notice that.
Re: neuroStock & neuroGrape mods by Andrey Kotlarski
Posted: Tue Apr 24, 2012 12:51 am
by Jim Ablett
gaard wrote:I am not getting the output logs using mode 0. Could this be because I am using Windows and you compiled with the original directory format "./logs/stockfish-"?
Windows builds are not outputting the neccesary log files.
I have re-compiled both NeuroStock & NeuroGrape with Cygwin and now both are working correctly
http://dl.dropbox.com/u/5047625/neurogrape-11-ja.zip
Mirror:
http://cl.ly/G4l7/neurogrape-11-ja.zip
http://dl.dropbox.com/u/5047625/neurostock-201-ja.zip
Mirror:
http://cl.ly/G5Rg/neurostock-201-ja.zip
Downside is that NeuroStock can only be 32 bit (Cygwin has no 64 bit), but I don't think it matters much with these experimental programs.
Jim.