Author: Peter Fendrich
Date: 09:50:23 06/28/04
Go up one level in this thread
I'm doing similar things with GA right now. The difference between GA and Simulated Annealing is probably of not very important here. My approach is somewhat different and I can elaborate more about it later on. One problem that you are probably aware of is the low number of games. I belive that small changes in an evaluation term will give minor differences in performance and 10 games will maybe not capture that. For instance, play a match vs some engine. Make a change that you know is lowering the strength and play the same match again maybe x times and see what you get. I would suggest shorter timecontrols in order to get more games. /Peter On June 28, 2004 at 08:54:00, Anthony Cozzie wrote: >Most of you probably know by now that my "secret project" with Bob has been >evaluation autotuning. I feel that with the improvements in parallel search and >CPU speed, the evaluation is beginning to be become more important than the >search in top programs (e.g. Shredder). As the number of terms increases, the >number of relations increases quadratically and the problem becomes harder. I >don't think that autotuning will ever be as good as hand tuning, but it is a >good way to get a first guess for the value of a parameter, as well as seeing >whether a new parameter is worth anything. This weekend, aside from watching >massive amounts of "Hajime no Ippo", I have finished a run with the latest >tweaks to the algorithm. > >The basic idea (which actually dates to my senior year in college, when I took >CAD tools, 18760) is to use simulated annealing. Simulated annealing makes >random changes and accepts them probablistically: > >(V1 > V0) : accept change >(V1 < V0) : accept if rand() > exp((V1 - V0)/T) > >Where T is the temperature, and it gradually gets smaller during the anneal. >This has the effect of ignoring small changes initially, and concentrating on >the big picture. This is the best link I can find in a few minutes of google >search; I will explain more if people are still unclear: > >http://members.aol.com/btluke/simann1.htm > >I am looking for some testers to run games with the latest crafty against >various engines. Basically, run N games against challenger X with the standard >settings, and then N games with the new settings. I am only really interested >in longer timecontrols: 20 min + on an Athlon 2.0G or so (70 min on P-650, etc), >and at least 10 games. If this is ever published, I will include everyone who >runs some test games in the acknowledgements section. > >anthony > >The crafty evaluation script: > >evaluation 1 100 >evaluation 2 300 >evaluation 3 300 >evaluation 4 500 >evaluation 5 900 >evaluation 11 100 >evaluation 12 0 >evaluation 13 100 >evaluation 14 100 >evaluation 15 100 >evaluation 16 100 >evaluation 17 120 >evaluation 21 10 >evaluation 22 3 >evaluation 23 16 >evaluation 24 1 >evaluation 25 50 >evaluation 26 8 >evaluation 27 13 >evaluation 28 525 >evaluation 29 150 >evaluation 30 50 >evaluation 31 100 >evaluation 32 10 10 10 10 10 10 10 10 9 22 -2 -11 21 50 15 10 13 20 16 17 26 21 >25 16 7 4 3 11 12 16 10 6 -1 1 7 11 7 13 8 0 -1 -1 4 9 4 6 5 0 -3 -5 -5 -7 -9 4 >4 -2 -3 -3 -3 -3 -3 -3 -3 -3 >evaluation 33 0 2 2 2 23 43 212 212 >evaluation 34 0 0 0 0 0 0 7 7 >evaluation 35 0 1 2 12 23 53 117 117 >evaluation 36 0 6 6 10 13 22 43 43 >evaluation 37 0 5 15 36 49 55 63 63 63 >evaluation 38 0 9 28 44 68 89 89 89 89 >evaluation 39 0 3 9 9 9 9 9 >evaluation 40 0 0 6 24 40 64 80 94 96 >evaluation 41 0 0 0 0 12 60 100 0 >evaluation 42 45 41 20 19 18 13 13 9 9 9 9 9 6 6 6 6 5 4 3 3 3 3 3 3 3 3 3 3 3 >3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 >3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 >3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 0 >evaluation 43 72 36 36 36 34 31 30 30 27 27 24 24 21 21 18 18 15 15 12 12 10 10 >10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 >10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 >10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 >10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 >evaluation 51 6 6 4 3 1 0 0 0 >evaluation 52 -32 -16 -8 -8 -8 -8 -16 -32 -24 -24 4 4 4 4 -24 -24 -8 2 6 6 6 6 >2 -8 -8 2 6 6 6 6 2 -8 -8 2 4 4 4 4 2 -8 -8 2 2 2 2 2 2 -8 -8 -8 0 0 0 0 -8 -8 >-8 -8 -8 -8 -8 -8 -8 -8 >evaluation 53 0 0 0 0 0 0 0 0 0 0 6 12 12 6 0 0 0 0 12 18 18 12 0 0 0 0 12 15 >15 12 0 0 0 0 0 6 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 >evaluation 61 10 >evaluation 62 2 >evaluation 63 12 >evaluation 64 4 >evaluation 65 175 >evaluation 66 -12 -18 6 19 18 22 11 -2 -13 >evaluation 67 7 7 6 3 1 0 0 0 >evaluation 68 -16 -16 0 0 0 0 -16 -16 -16 2 4 4 4 4 2 -16 -4 2 6 6 6 6 2 -4 -4 >2 6 6 6 6 2 -4 -4 2 4 4 4 4 2 -4 -4 2 2 2 2 2 2 -4 -16 2 0 0 0 0 2 -16 -16 -16 >-8 -8 -8 -8 -16 -16 >evaluation 71 16 >evaluation 72 3 >evaluation 73 4 >evaluation 74 9 >evaluation 75 17 >evaluation 76 10 >evaluation 77 20 >evaluation 78 23 >evaluation 79 10 >evaluation 80 9 9 6 4 2 0 0 0 >evaluation 81 6 6 2 1 1 0 0 0 >evaluation 82 0 0 4 6 6 4 0 0 0 0 4 6 6 4 0 0 0 0 4 6 6 4 0 0 0 0 4 6 6 4 0 0 0 >0 4 6 6 4 0 0 0 0 4 6 6 4 0 0 0 0 4 6 6 4 0 0 0 0 4 6 6 4 0 0 >evaluation 91 22 >evaluation 92 6 >evaluation 93 50 >evaluation 94 6 >evaluation 95 8 >evaluation 96 10 10 10 4 0 0 0 0 >evaluation 97 9 9 5 0 0 0 0 0 >evaluation 98 0 0 0 0 0 0 0 0 -15 0 4 5 5 4 0 -15 0 2 4 10 10 4 2 0 0 2 10 12 >12 10 2 0 -10 2 10 12 12 10 2 -10 -10 -10 4 10 10 4 -10 -10 -10 2 8 8 8 8 2 -10 >-10 -8 0 0 0 0 -8 -10 >evaluation 101 15 >evaluation 102 96 >evaluation 103 600 >evaluation 104 7 >evaluation 105 -86 >evaluation 106 3 >evaluation 107 0 0 0 0 0 0 0 0 1 1 1 2 2 4 4 4 4 7 11 13 13 18 20 20 27 27 30 37 >37 37 37 37 37 43 43 43 47 47 47 47 53 54 57 59 61 61 61 61 61 61 64 65 65 65 66 >68 69 72 72 75 80 80 87 87 97 98 99 108 154 165 184 217 245 249 286 397 426 553 >689 799 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 >922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 922 >922 922 922 922 922 922 922 922 922 922 >evaluation 108 8 8 8 8 8 8 8 8 7 7 7 7 7 7 7 7 6 6 6 6 6 6 6 6 5 5 5 5 5 5 5 5 4 >4 4 4 4 4 4 4 3 3 3 3 3 3 3 3 1 1 2 3 3 2 1 1 1 0 2 3 3 2 0 1 >evaluation 109 -50 -30 -30 -30 -30 -30 -30 -50 -30 -30 0 0 0 0 -30 -30 -30 -10 >30 40 40 30 -10 -30 -30 -10 30 40 40 30 -10 -30 -30 -10 20 30 30 20 -10 -30 -30 >-10 0 10 10 0 -10 -30 -30 -20 -10 0 0 -10 -20 -30 -50 -40 -30 -20 -20 -30 -40 >-50 >evaluation 110 -90 -70 -50 -30 -30 -30 -30 -30 -70 -50 -30 -30 0 0 0 -10 -70 -50 >-30 -10 30 30 30 0 -70 -50 -30 -10 30 30 30 0 -70 -50 -30 -10 30 30 30 0 -70 -50 >-30 -10 20 20 20 0 -70 -50 -30 -20 -10 -10 -10 -10 -90 -70 -50 -40 -30 -20 -20 >-30 >evaluation 111 -30 -30 -30 -30 -30 -50 -70 -90 -10 0 0 0 -20 -30 -50 -70 0 30 30 >30 -10 -30 -50 -70 0 30 30 30 -10 -30 -50 -70 0 30 30 30 -10 -30 -50 -70 0 20 20 >20 -10 -30 -50 -70 -10 -10 -10 -10 -20 -30 -50 -70 -30 -20 -20 -30 -40 -50 -70 >-90 >evaluation 112 0 3 15 16 >evaluation 113 0 12 13 16 >evaluation 114 0 4 13 20 21 25 31 35 35 39 40 40 40 48 52 56 56 56 56 56 56 56 >56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 >56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 >evaluation 115 16 16 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 24 25 25 >26 26 27 27 28 28 29 29 30 30 31 31 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 >32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 >evaluation 121 10 >evaluation 122 7 >evaluation 123 12 >evaluation 124 20 >evaluation 125 2 >evaluation 126 3 >evaluation 127 20
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.