On 09/21/2011 10:00 PM, Michael Petch wrote: On 21/09/2011 12:31 PM, Johannes Bickmann wrote: Hello list! I'm a new member in this list. I recognized that the md5-sums of my generated bearoff databas
Those files I believe are slightly different than the form used now so they won't match what is produced locally by you. Usually you can tell if the database and neural net files are loaded properly
What if we combine the two approaches? 1) Build a OSR database using the mean and std 2) Compare this to the full database that you are generating and for each position, generate a metric on how good
Regarding bearoff databases, GNUBG supports only 1 two sided bearoffdatabase and 1 one-sided bearoffdatabase at any given time. As for the file naming convention and other info that may help you, y
Hi all, just noticed this, in the "Evaluation engine" details (Help/About/Evaluation engine): * Contact neural network evaluator: - version 0.15, 250 inputs, 128 hidden units. * Crashed neural networ
I suppose it helps that the opponent's position is a few-rolls position and more balanced long races would not do as well. Yes! There it actually two ways to do the dynamic programming of two sided
I suppose it helps that the opponent's position is a few-rolls position and more balanced long races would not do as well. Yes! There it actually two ways to do the dynamic programming of two sided
Frank, Hugh Sconyers "BEAROFF EQUITIES AND BACKGAME PROBABILITIES" [1] is a commerical product ($100) and consists of 12 DVD. It is not related to the database you downloaded, which in no way can r
I think we discussed this a while ago: cubeless winning chances in match play are taken from the 2s dB, janowski formula converts them into cubefull then plies are applied. Hence cubeless winning cha
There has been a significant amount of discussion on BGO about this position of Neil¹s. http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=47213 Position ID: gwEAQCoAAAAAAA Match ID: cAmg
hi I have some new info. if I use eval threads = 1, the match finishes without any problem. if I use eval threads = 2, the gnu crashes with error message below. this is on a dual processor intel 930
The most common reason is that the bearoffdatabase has been downloaded onto a Windows machine as an ASCII rather than a binary file. The will produce a corrupt file (every occurrence of a byte with
I don't think makebearoff will successfully create files larger than 6 * 11, although it's not the file size itself that is the problem, as I understand it. When I tried it (years ago now), it ran in
Unfortunately there are 10M positions with 13 chequers on 13 points, and 15 positions with 2 chequers on five points, hence a total of 150M positions. I keep the mean and the variance for the roll di
Before beginning. 1) Go to Settings menu/Players. Set Player 0 as GNU Backgammon at and Expert Level. Player 1 should be human 2) Go to Settings/Analysis. Uncheck "Same as Analysis" and select Expert
Hello Ian, thank you for your answer. A nice catalogue of questions. This is the style of constructive discussion I was looking for. A short preamble to the history of my work. I decided to improve
No magic involved here. getBearoffGammonProbs takes the position of the side wanting to bear one checker. (Since it is a bearoff, all other 19 points are know to be empty, so only 6 are needed). It