Re: [Computer-go] Regarding MCTS playouts

2017-04-05 Thread Michail Drozdov
Dear Alvaro, Thank you, that alone sped up playouts more than 2 times. On 5 Apr 2017 6:12 p.m., "Álvaro Begué" wrote: For identifying points that look like eyes, it's useful to have a 16-bit value at each position of the board that contains the colors of the 8 neighbors

[Computer-go] Google TPU performance

2017-04-05 Thread Peter Kollarik
i just found this: https://drive.google.com/file/d/0Bx4hafXDDq2EMzRNcy1vSUxtcEk/view Peter ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Aya's selfplay games for training value network

2017-04-05 Thread Álvaro Begué
Oh, that's unfortunate. When I read the description of what you did, I thought the whole point of using of a sigmoid in the last stage was precisely so the ownership map would be reasonably correct. Your database should still be very useful, but it feels like it should be much easier for the DCNN

Re: [Computer-go] Regarding MCTS playouts

2017-04-05 Thread Álvaro Begué
For identifying points that look like eyes, it's useful to have a 16-bit value at each position of the board that contains the colors of the 8 neighbors (2 bits per neighbor, with an encoding like 00=empty, 01=black, 10=white, 11=outside). You can maintain this incrementally when a point on the

[Computer-go] Regarding MCTS playouts

2017-04-05 Thread Michail Drozdov
Hello, I'm trying to write my own go playing program. Right now it is GnuGo style program - no dedicated search is performed. It is pleasure to see bot improving, but recently I decided to learn and implement MCTS, meaning I have to use my go board implementation often (for playouts). I see some

Re: [Computer-go] Aya's selfplay games for training value network

2017-04-05 Thread Lukas van de Wiel
Hi Yamashita Thanks you for uploading them, and explaining how they are organised. Your contribution is much appreciated! Lukas On 4/5/17, Hiroshi Yamashita wrote: > Hi, > > I have uploaded Aya's selfplay games. > There are about 1,440,000 games. > Selfplay program strength

[Computer-go] Aya's selfplay games for training value network

2017-04-05 Thread Hiroshi Yamashita
Hi, I have uploaded Aya's selfplay games. There are about 1,440,000 games. Selfplay program strength is about KGS 4d and 5d. Aya's selfplay games for training value network http://www.yss-aya.com/ayaself/ayaself.html Using these games, KGS 4d over, tygem 9d, and GoGoD, I made slightly (+70