Confident Local Optimization for Noisy Black-Box Parameter Tuning
Author: Rémi Coulom

Artificial intelligence in games often leads to the problem of parameter tuning. Some heuristics may have coefficients, and they should be tuned to maximize the win rate of the program. A possible approach is to build local quadratic models of the win rate as a function of program parameters. Many local regression algorithms have already been proposed for this task, but they are usually not robust enough to deal automatically and efficiently with very noisy outputs and non-negative Hessians.

The CLOP principle
- a new approach to local regression that overcomes all these problems in a simple and efficient way.

CLOP-0.0.9.tar.bz2 - source code
CLOP-Win-0.0.9.zip - Windows version

Links and External Contributions
CCC discussion of CLOP
computer-go discussion of CLOP

Experiments and Data
PieceValues.tar.bz2 - a data file by Gian-Carlo Pascutto demonstrating the application of CLOP to piece values in Chess.

Connection Scripts
cutechess-cli has support for CLOP.
chessScript.02.tar.gz - a script to connect an UCI chess engine to CLOP, by Don Dailey (new version updated 2011-09-08).
Supportfor CLOP

These are some screenshots of an old version in action. You can also run the program from the command line, which is more convenient for use on a remote cluster. The program can deal with chess outcomes (win/draw/loss), and integer parameters. The program is written in C++ with Qt, so it can be compiled and run on Windows, Linux, and MacOS.
[Hide][You must be registered and logged in to see this image.]