Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fruit fly races

Author: Michael Yee

Date: 05:14:22 04/07/05

Go up one level in this thread


Hi Anthony,

I don't really understand your strong words, since there are at least a few
examples I can think of where very different algorithms can co-exist.

For example, in linear programming, there's the simplex method (moving from
extreme point to extreme point in a discrete fashion) and there are interior
point methods (starting from an interior point of the feasible region and using
a nonlinear technique [newton step] to move toward the optimal solution). Both
algorithms perform well in practice.

And (perhaps more relevant to the chess discussion), for solving hard
optimization problems there are some "standard" techniques (say, using branch
and bound and using LP or other relaxations to get bounds) and some nonstandard
ones (e.g., meta-heuristics). I'm pretty sure you're familiar with these
already, but approaches like genetic algorithms, simulated annealing, tabu
search, ant colonies, etc., all get their inspiration from natural systems that
seem to be able to efficiently solve their particular optimization problems. So
both math programming based approaches and nature-insipired approaches can be
viable though very different.

It's not clear to me why chess has only *one* family of good algorithms.

Michael



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.