Author: Bob Durrett
Date: 15:07:35 11/18/00
Go up one level in this thread
On November 18, 2000 at 17:52:21, Graham Laight wrote: >On November 18, 2000 at 17:37:03, Bob Durrett wrote: > >>I read somewhere that the human brain stores pattern information in a manner at >>least vaguely similar to the way a Fourier Transform works. >> >>For an elementary example, one could store a single-variable function, f(t) as a >>spectrum where amplitude and phase are both stored. >> >>For information to be stored in the human brain, the "functions" are not so >>simple. >> >>But, with a stretch of the imagination, one might see how this could be. >> >>The key idea is that the information is stored in a large percentage of the >>brain's memory, so that if one were to accidentally [or intentionally] remove >>part of the brain, the memory could still be recoverable, though not so >>precisely. This assumes a lot, of course, like not killing the human, etc., but >>consider the principle. >> >>Now, look at the modern computer's memory. What if information could be >>transformed in some manner so that the information would be stored in a >>distributed manner in the computer's memory? Maybe not practical today [???] >>but I don't see any reason why someone could not figure out how to do that. >> >>Well, to be more specific, suppose one wishes to store 10,000 chess positions. >>Then, the code would have to be configured [by some really smart programmer] so >>that it would compare the current game position to the stored memory. It would >>be a matter of correlation. The stored pattern which gave the highest >>correlation [after doing the equivalent of an inverse Fourier Transform] would >>subsequently be recalled and used by the code to figure out what move, or chess >>idea or plan, to give priority too. Incidentally, not only would the 10,000 >>positions be stored, but also the information as to how to proceed in that >>position. >> >>Maybe it would require more of a computer than we currently have. [an >>understatement?] >> >>But, if you're game, think about it! > >Fourier transforms are usually applied to waves, to determine the extent to >which sine waves exist in them. Are you saying that they could also be applied >to chess positions? If so, please explain how. > >-g Well, it takes some willingness to think about functions which are not so elementary as "waves." Your "wave" is a single-valued real function of a real variable, time. But multi-dimensional Fourier transforms are also commonplace for multi-dimensional continuous functions, even of multi-dimensional independent variables. Then one can extend this to digitized versions. Hence we start with an elementary digital fourier transform ["FFTs"]and then extend to the multi-dimensional cases. So far, all "Ho-hum" everyday commonplace. Two dimensional Fourier transforms are used in optics. [Holograms, etc.] There is no reason why a chess position could not be transformed inasmuch as it is two-dimensional. But this is too elementary an example. It is not necessary to have a two-dimensional photograph to store a chess position. That would be a terribly inefficient way to store the esential information. So, if the position information is stored in some "suitable" digital format, then a "really smart" person [someone of Fourier's calibre] should be able to figure out a suitable "digital transform" which would convert the information into a "transformed" state suitable for storage over a reasonably large part of memory. It would not necessarily have to be "Fourier." Since computer memories are just 0s and 1s, one must "think digital." That stretches one's imagination a little bit, but problems are made to be solved. This is just another detail to be worked out. One essential idea I may not have made clear is that you can store MANY positions in the SAME block of memory. MIND STRETCH
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.