Can a neural network predict results based on patterns that exist in its training set

Question

This may sound like an open question, and I apologize in advance. I know that a neural network can predict the outcome based on patterns found in the training dataset. Can a neural network predict results based on models that it has not observed. If not, which algorithm should be used to solve this problem.

Answer 1

Most neural network applications operate on the basis of supervised training, which means that the algorithm receives feedback on its performance. The algorithm uses feedback to adjust the weight of the network. The most common mechanism for this setting is backpropagation, but there are others.

Training Set vs Training Signal.

Your question mentions the training kit. However, the fact that neural networks require feedback does not necessarily imply a training set; this is just one alternative. Sometimes feedback can sometimes be found directly from the environment in which the network operates.

Examples of feedback without training set:

(1) , . , , .

(2) , . . OCR.

, , . , ( ).

Hebbian.

Hebbian. Hebb:

A , B , , , , B, .

", , ". , --. .

.

, , . , . , n , , - , : . " ". , .

, , . , , , . - , .

, , . (SOFM), . SOFM .

. . (Attribution: By Denoir - , CC BY-SA 3.0, https://en.wikipedia.org/w/index.php?curid=40452073)

SOFM .

, , SOFM . , (. .).

Answer 2

, , , , , . , , . , :

• :, . , : : . - , - , - , - , , . , , . , , , , . , . , , . , , , - , .
• :. , .

, , . , - .