Author: Wylie Garvin
Date: 10:05:34 12/30/01
Go up one level in this thread
On December 29, 2001 at 05:11:26, Tom Kerrigan wrote: >Here's my idea. > >You have a position and you want your program to play a certain move (which it >presumably isn't playing). You run this algorithm: > >1. Search the position, get a PV. The evaluation of the last position of the PV >is eval(1). >2. Search only the move that you want your program to make, get a PV. This >end-point evaluation is eval(2). >3. Figure out which eval terms are different between eval(1) and eval(2). >Decrease the weights of all the different eval(1) terms slightly. Increase the >eval(2) terms slightly. >4. Repeat until the program plays the move you want. > >You could run this on lots of positions from GM games, to get your program to >play like a GM. (At least in some positions, heh.) > >Has this been done before? Are there any glaring problems with this idea? Does >anybody want to try this? If so, I'd like some credit for it. If not, I'll >probably get around to trying it sometime... > >-Tom Hi Tom, A lot of people have replied already, but just in case no one has mentioned it you should check out Baxter et. al.'s TDLeaf() algorithm. TDLeaf(): Combining Temporal Difference Learning with Game-Tree Search http://citeseer.nj.nec.com/139694.html They trained their program's eval function automatically on a large set of grandmaster games. I did some fooling around with this once with my AI prof, his idea was, rather than adjust the evaluation incrementally, to collect a long list of constraints (one for each position in which the program would play the "wrong" move) and then simultaneously solve as many of the constraints as possible. But we didn't pursue it too far because it turned out that Baxter et. al. had made a few clever optimizations in their program (I think it was KnightCap they were working with, if I remember right) which gave it most of the benefits of optimizing globally. Wylie
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.