Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A final stab at big-O

Author: Dann Corbit

Date: 17:18:09 05/09/01

Go up one level in this thread


On May 09, 2001 at 20:07:29, Ricardo Gibert wrote:

>On May 09, 2001 at 18:28:21, Dann Corbit wrote:
>
>>On May 09, 2001 at 18:20:59, Ricardo Gibert wrote:
>>
>>>On May 09, 2001 at 17:40:16, Dann Corbit wrote:
>>>
>>>>On May 09, 2001 at 17:32:27, Ricardo Gibert wrote:
>>>>[snip]
>>>>References: A. S. Fraenkel and D. Lichtenstein, Computing a perfect strategy for
>>>>n*n chess requires time exponential in n, Proc. 8th Int. Coll. Automata,
>>>>Languages, and Programming, Springer LNCS 115 (1981) 278-293 and J. Comb. Th. A
>>>>31 (1981) 199-214.
>>>>
>>>>FCOL
>>>
>>>n is clearly finite, but unbounded in the above, so the time required is
>>>exponential. All quite correct. This is a perfect example that makes *my* case.
>>>The authors clearly understand about big-O perfectly. They took the pains to
>>>generalize the problem to n*n chess. They didn't just say chess, because they
>>>understood perfectly that that would not be correct, which is all that I've been
>>>saying.
>>>
>>>It's clear people cannot tell when a variable is bounded or unbounded and what
>>>is meant by finite and infinite and when a variable has been instantiated and
>>>when it has not and what the difference between an instantiated variable and a
>>>constant, etc.
>>
>>Here are some references that classify chess as NP.  One goes so far as to offer
>>a formal proof:
>>
>
>The references to chess in this one are incoherent:
>>http://www.cs.duke.edu/~dr/1.00s/L37.html

"Intractable Algorithms

Chess
Note: here N is number of moves looking ahead.
We Have an Algorithm!
Layers of look ahead: If I do this, then he does this, ....
Problem Solved (?!)
Can Represent Possibilities by Tree
Assume 10 Possibilities Each Move
t = A * 10^N
!!!!
Exponential
Sample Numbers "

Seems plain enough to me.


>This one references "linear chess" but does not define what they mean by it:
>>http://www.cs.bu.edu/fac/lnd/toc/z/node18.html
>
>Here they discuss alphabeta and minimax as applied to chess. Just because
>alphabets and minimax are O(exp(n)) does not mean that task of solving chess in
>O(exp(n)):
>>http://www.glyph.pe.kr/phoenix/AI/AIChess4.htm
>
>This next one has the following to say about chess:
>
>"Exhaustive searching and naive game-playing programs (e.g., chess) have
>exponential time complexity: O(2^N)."
>
>They qualify what they say with the word "naive", so I can't say they're wrong:
>>http://www.cis.ksu.edu/~schmidt/300f00/Lectures/Week3.html
>
>The following is a less readable version of 2nd one you give above, which leads
>me conclusion that you did not really examine this nor any of the others with
>any care:
>>http://www.cs.bu.edu/fac/lnd/toc/n-npcom
>>
>>Are they all confused?
>>Or are they simply using the standard definition of P/NP?
>
>Note:
>(1) that a "game playing program" is not the same as "solving chess" a program
>that plays random moves is clearly not O(exp(n)).

No arguments there.  I do submit (however) that the task of solving additional
plies is exponential in difficulty, even from a random board position.  At least
using my definitions.

>(2) saying that "a certain algorithm that solves chess is O(f(n))" is not the
>same as saying "solving chess is O(f(n))". You yourself gave an algorithm that
>sorts in O(n!), even though the problem of sorting is O(n*log(n)) or even less.

No arguments there either.

I think I understand your definition well enough.  It's different than the one I
use, but well -- there we are.

I use my definition to calculate whether tasks are feasible or not.  Probably
because my degree is in applied mathematics (numerical analysis) and I suspect
your degree is in pure maths of some sort.

The definition I use is valuable to me because it helps me to get work done
effectively.  I cannot use the definition that you offer because it does not
create pragmatic or heuristic results.

So I think it's time to simply give up the ghost as far as trying to reach some
conclusion.

I am certainly aware that there are others who use a definition very different
from the one I use.  As long as they don't use it to perform feasibility
studies, I don't see any real harm in that.



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.