Does anybody know about back propagation algoritms and how implements it into program if one of you know about it and see the good tutorial just tell me please ?

This is neural network algoritms that can be used for inteligence programming for computer.

let me take one example a child can recognize a picture fastest than computer done. But child is very difficult to do mathemathical processing like what is the sum of 3 + 3 ? compare it with computer done to do calculation with a million data in a short time.

i need this program by simple example may be in this forum i can find who really expert in this topic.

all response will be appreciated.
Posted on 2001-07-24 07:03:07 by newbies
Recognizing a an image is done by comparison, mainly by the outline or by evidence of a detail. Remembering is a lengthy process even to humans, but a little detail is the spot where we usually work on.

The main fault of electronic (or artificial) intelligence is the way how this detail is stored, and which detail is chosen. For example, an image taken from nature is made mainly of colors, not of water, leaves, objects. Remembering a forest picture therefore is done by green, and searching for green pictures is easier and faster than for searching for a tree.

But still, AI is not comparable to human's intelligence, and the best formula for AI is: competence, limit and priority, all of them easily implementable in a multithreaded operating system.

It's just useless to ask a disk driver to remember a picture, since it isn't it's competence. Asking a search algo more than the database contains, is useless, because there is a limit, and asking for an information while the proper artificial live is in danger, is absolutely to avoid, because there are priorities.

All you need are absolutely indipendent objects which know these three terms on their own, and know how to interact with each other. Not easy, but possible, mainly with self modifying algo's. If you start with little, you may have surprising results, just dedicating tiny objects to tiny jobs, with a short lifetime :)

I'd like to point out DIV2, which is a development environment for games and has the right concept of those objects, and on a very <simple> base it is capable to learn and to interact with other objects. Have fun,

Giovanni
Posted on 2001-08-11 10:21:35 by sch.jnn
Back porpagation is not an algoratin, its a method of supervised training Neural Nets.

Its the nets that are used for intelligence and depending on how the net is written will decide how and even if back propagation can be implemented, for example a net built on the same structre as the human brain cannot learn by back propagation.
Posted on 2001-08-11 11:17:01 by Eóin
To recognize vowel i must know about Formant frequency. Does anybody know about it ? where is the site contain a good tutorial about that topic ? with example is preferable :->

all response will be appreciated
Posted on 2001-08-23 05:06:05 by newbies