neophyte
Senior Heliman
Location: California
| Gentlemen:
Please correct me if I am wrong. A neural network, implemented by software on a PC, consists of the following:
A matrix (as in a big tac-tac-toe pattern, that has columns and rows) that recieves and records a number of patterns.
As the patterns are recieved, they are also assigned a 'good' or 'bad' designation, by the teacher. (training)
If the pattern is 'good' it is remembered and reinforced. If the pattern is 'bad', it is disregarded, and lost.
Each new pattern is evaluated in view of the patterns already judged as 'good', and the designation given by the teacher for this particular pattern as good, or bad.
Over time, and many repetitions, the matrix comes to descriminate the good patterns with more detail. (learning)
It is thought that this is the way biological brains work. Hence, the name neural networks. A neuron is a single cell, with many connections, and the capacity to repeat that which is familiar.
Computers, on the other hand, operate on a completely differant idea. Computers do one thing at a time. Computers take the input signal, (ones and zeroes) and process it according to their software, then output the result (more ones and zeros). Nothing more, nothing less. Computers do not 'learn' with conventional software. They just 'do'.
Is this forum talking about 'Neural Networks' that learn? or just parellel processors? You be the judge. Or correct me, if you think I am wrong. I have been wrong before!!! |
