]> git.sesse.net Git - wloh/log
wloh
12 years agoAdd a prior, to smooth out the mu a bit.
Steinar H. Gunderson [Sat, 17 Mar 2012 13:02:27 +0000 (14:02 +0100)]
Add a prior, to smooth out the mu a bit.

12 years agoSkip 0-150 and 150-0 matches for the purpose of the rating.
Steinar H. Gunderson [Sat, 17 Mar 2012 13:02:13 +0000 (14:02 +0100)]
Skip 0-150 and 150-0 matches for the purpose of the rating.

12 years agoProperly unlink the temporary file after training.
Steinar H. Gunderson [Sat, 17 Mar 2012 12:10:06 +0000 (13:10 +0100)]
Properly unlink the temporary file after training.

12 years agoAdd a script to sync all data from the master WLoH database.
Steinar H. Gunderson [Sat, 17 Mar 2012 12:09:30 +0000 (13:09 +0100)]
Add a script to sync all data from the master WLoH database.

12 years agoAdd a script to train the model.
Steinar H. Gunderson [Sat, 17 Mar 2012 12:07:33 +0000 (13:07 +0100)]
Add a script to train the model.

12 years agoMake the output more machine-friendly.
Steinar H. Gunderson [Sat, 17 Mar 2012 12:03:32 +0000 (13:03 +0100)]
Make the output more machine-friendly.

12 years agoDon't print out Hessian by default.
Steinar H. Gunderson [Sat, 17 Mar 2012 01:30:58 +0000 (02:30 +0100)]
Don't print out Hessian by default.

12 years agoTake weight into account in Hessian.
Steinar H. Gunderson [Sat, 17 Mar 2012 01:30:45 +0000 (02:30 +0100)]
Take weight into account in Hessian.

12 years agoCompute the Hessian instead of the Fisher information matrix. The numerics here suck.
Steinar H. Gunderson [Sat, 17 Mar 2012 01:29:41 +0000 (02:29 +0100)]
Compute the Hessian instead of the Fisher information matrix. The numerics here suck.

12 years agoOptimize global sigma.
Steinar H. Gunderson [Sat, 17 Mar 2012 00:12:59 +0000 (01:12 +0100)]
Optimize global sigma.

12 years agoAdd weight to all matches, and optimize mu based on it.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:52:03 +0000 (00:52 +0100)]
Add weight to all matches, and optimize mu based on it.

12 years agoUse sum-of-squares instead absolute-single-greatest-change as metric for convergence.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:31:49 +0000 (00:31 +0100)]
Use sum-of-squares instead absolute-single-greatest-change as metric for convergence.

12 years agosizeof() should be concrete variables when possible.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:29:53 +0000 (00:29 +0100)]
sizeof() should be concrete variables when possible.

12 years agoIncrease the max number of iterations a bit.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:29:17 +0000 (00:29 +0100)]
Increase the max number of iterations a bit.

12 years agoOnly dump scores once, at the very end.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:28:49 +0000 (00:28 +0100)]
Only dump scores once, at the very end.

12 years agoMore sensible dumping.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:26:42 +0000 (00:26 +0100)]
More sensible dumping.

12 years agoIncrease MAX_PLAYERS a lot.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:24:36 +0000 (00:24 +0100)]
Increase MAX_PLAYERS a lot.

12 years agoInitial commit.
Steinar H. Gunderson [Fri, 16 Mar 2012 23:23:05 +0000 (00:23 +0100)]
Initial commit.