Author: Gareth McCaughan
Date: 13:56:01 11/27/01
Go up one level in this thread
On November 27, 2001 at 12:51:52, Dann Corbit wrote: > Here is a post in the Winboard forum with some explanation: > http://f11.parsimony.net/forum16635/messages/18447.htm (Summary: Dann's optimizing F(parameters) = performance on a bunch of test suite positions by taking 1-d slices through parameter space parallel to the coordinate planes and looking for a minimum along each slice.) > There may be a sort of simulated annealing approach. Clearly, the parameters > may interact. However, as I iterate through the parameters from one depth to > the next, I keep the minimum calculated for the previous parameter. Why don't you use a slightly more sophisticated optimization algorithm? Say, the quadratically convergent algorithm of Powell described in *ick* Numerical Recipes? Cool experiment, anyway. -- g
This page took 0 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.